Inference Scaling Evaluation Results Batch Conversion

data/frontiermath/anthropic/claude-opus-4-20250514/0467b4f5-90b0-40f9-9119-6dfd78865ad4.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786111930.465773",
+    "evaluation_timestamp": "1775558527.0",
+    "retrieved_timestamp": "1786111930.465773",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+8ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775558527.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.1875,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.11967838846954228
+                    },
+                    "num_samples": 48
+                },
+                "details": {
+                    "total_matched_trajectories": "18",
+                    "stopping_reason_count_repetition_guard": "17",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "500251.61",
+                    "total_total_tokens_target_model": "9004529.0",
+                    "avg_total_tokens_other_models": "110.50",
+                    "total_total_tokens_other_models": "1989.0",
+                    "avg_total_tokens_all_models": "500362.11",
+                    "total_total_tokens_all_models": "9006518.0",
+                    "avg_turn_count": "23.83",
+                    "total_turn_count": "429.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/0467b4f5-90b0-40f9-9119-6dfd78865ad4_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "95fc5fd9c0a116e8d44e8bb7ec6ee2531974973ff72a96f2155c34ac2f24adbf",
+        "total_rows": 48
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/0467b4f5-90b0-40f9-9119-6dfd78865ad4_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95fc5fd9c0a116e8d44e8bb7ec6ee2531974973ff72a96f2155c34ac2f24adbf
+size 8134100

data/frontiermath/anthropic/claude-opus-4-20250514/3433946a-b956-48a1-bb09-14a8f82ec93f.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786113485.910628",
+    "evaluation_timestamp": "1775558527.0",
+    "retrieved_timestamp": "1786113485.910628",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+7ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775558527.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.14285714285714282,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.06388765649999399
+                    },
+                    "num_samples": 42
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "10",
+                    "stopping_reason_count_completed_on_successful_submit": "2",
+                    "avg_total_tokens_target_model": "377446.83",
+                    "total_total_tokens_target_model": "4529362.0",
+                    "avg_total_tokens_other_models": "97.50",
+                    "total_total_tokens_other_models": "1170.0",
+                    "avg_total_tokens_all_models": "377544.33",
+                    "total_total_tokens_all_models": "4530532.0",
+                    "avg_turn_count": "22.58",
+                    "total_turn_count": "271.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/3433946a-b956-48a1-bb09-14a8f82ec93f_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb",
+        "total_rows": 42
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/3433946a-b956-48a1-bb09-14a8f82ec93f_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb
+size 6162708

data/frontiermath/anthropic/claude-opus-4-20250514/3491820d-0d2f-4811-b970-f186b857acf5.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786111879.107692",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786111879.107692",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+3ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.05555555555555555,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.05555555555555556
+                    },
+                    "num_samples": 18
+                },
+                "details": {
+                    "total_matched_trajectories": "18",
+                    "stopping_reason_count_repetition_guard": "17",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "453550.72",
+                    "total_total_tokens_target_model": "8163913.0",
+                    "avg_total_tokens_other_models": "110.50",
+                    "total_total_tokens_other_models": "1989.0",
+                    "avg_total_tokens_all_models": "453661.22",
+                    "total_total_tokens_all_models": "8165902.0",
+                    "avg_turn_count": "24.00",
+                    "total_turn_count": "432.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/3491820d-0d2f-4811-b970-f186b857acf5_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f",
+        "total_rows": 18
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/3491820d-0d2f-4811-b970-f186b857acf5_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f
+size 2812526

data/frontiermath/anthropic/claude-opus-4-20250514/7352a6c7-6dbe-480d-ae43-9e12bfb99ccd.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786117806.998392",
+    "evaluation_timestamp": "1775558527.0",
+    "retrieved_timestamp": "1786117806.998392",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+8ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775558527.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.1875,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.11967838846954228
+                    },
+                    "num_samples": 48
+                },
+                "details": {
+                    "total_matched_trajectories": "18",
+                    "stopping_reason_count_repetition_guard": "17",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "500251.61",
+                    "total_total_tokens_target_model": "9004529.0",
+                    "avg_total_tokens_other_models": "110.50",
+                    "total_total_tokens_other_models": "1989.0",
+                    "avg_total_tokens_all_models": "500362.11",
+                    "total_total_tokens_all_models": "9006518.0",
+                    "avg_turn_count": "23.83",
+                    "total_turn_count": "429.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/7352a6c7-6dbe-480d-ae43-9e12bfb99ccd_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "95fc5fd9c0a116e8d44e8bb7ec6ee2531974973ff72a96f2155c34ac2f24adbf",
+        "total_rows": 48
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/7352a6c7-6dbe-480d-ae43-9e12bfb99ccd_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95fc5fd9c0a116e8d44e8bb7ec6ee2531974973ff72a96f2155c34ac2f24adbf
+size 8134100

data/frontiermath/anthropic/claude-opus-4-20250514/7a63b33b-f568-4a1f-b5e1-2a8a00541d17.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786113468.298958",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786113468.298958",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+3ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.05555555555555555,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.05555555555555556
+                    },
+                    "num_samples": 18
+                },
+                "details": {
+                    "total_matched_trajectories": "18",
+                    "stopping_reason_count_repetition_guard": "17",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "453550.72",
+                    "total_total_tokens_target_model": "8163913.0",
+                    "avg_total_tokens_other_models": "110.50",
+                    "total_total_tokens_other_models": "1989.0",
+                    "avg_total_tokens_all_models": "453661.22",
+                    "total_total_tokens_all_models": "8165902.0",
+                    "avg_turn_count": "24.00",
+                    "total_turn_count": "432.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/7a63b33b-f568-4a1f-b5e1-2a8a00541d17_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f",
+        "total_rows": 18
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/7a63b33b-f568-4a1f-b5e1-2a8a00541d17_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f
+size 2812526

data/frontiermath/anthropic/claude-opus-4-20250514/a59f3717-c5cb-4f0e-88df-bf43049dcf62.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786111902.18625",
+    "evaluation_timestamp": "1775558527.0",
+    "retrieved_timestamp": "1786111902.18625",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+7ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775558527.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.14285714285714282,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.06388765649999399
+                    },
+                    "num_samples": 42
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "10",
+                    "stopping_reason_count_completed_on_successful_submit": "2",
+                    "avg_total_tokens_target_model": "377446.83",
+                    "total_total_tokens_target_model": "4529362.0",
+                    "avg_total_tokens_other_models": "97.50",
+                    "total_total_tokens_other_models": "1170.0",
+                    "avg_total_tokens_all_models": "377544.33",
+                    "total_total_tokens_all_models": "4530532.0",
+                    "avg_turn_count": "22.58",
+                    "total_turn_count": "271.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/a59f3717-c5cb-4f0e-88df-bf43049dcf62_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb",
+        "total_rows": 42
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/a59f3717-c5cb-4f0e-88df-bf43049dcf62_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb
+size 6162708

data/frontiermath/anthropic/claude-opus-4-20250514/c9887853-4943-4cd6-97a3-2f5820386272.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786117780.414865",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786117780.414865",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+3ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.05555555555555555,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.05555555555555556
+                    },
+                    "num_samples": 18
+                },
+                "details": {
+                    "total_matched_trajectories": "18",
+                    "stopping_reason_count_repetition_guard": "17",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "453550.72",
+                    "total_total_tokens_target_model": "8163913.0",
+                    "avg_total_tokens_other_models": "110.50",
+                    "total_total_tokens_other_models": "1989.0",
+                    "avg_total_tokens_all_models": "453661.22",
+                    "total_total_tokens_all_models": "8165902.0",
+                    "avg_turn_count": "24.00",
+                    "total_turn_count": "432.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/c9887853-4943-4cd6-97a3-2f5820386272_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f",
+        "total_rows": 18
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/c9887853-4943-4cd6-97a3-2f5820386272_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e264a14d305909abcc58391d372814256aa45512b23cb7038a6ee46581b8c95f
+size 2812526

data/frontiermath/anthropic/claude-opus-4-20250514/e5442618-79a9-40f6-83b5-95497d0bd979.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786117794.498366",
+    "evaluation_timestamp": "1775558527.0",
+    "retrieved_timestamp": "1786117794.498366",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+7ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775558527.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.14285714285714282,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.06388765649999399
+                    },
+                    "num_samples": 42
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "10",
+                    "stopping_reason_count_completed_on_successful_submit": "2",
+                    "avg_total_tokens_target_model": "377446.83",
+                    "total_total_tokens_target_model": "4529362.0",
+                    "avg_total_tokens_other_models": "97.50",
+                    "total_total_tokens_other_models": "1170.0",
+                    "avg_total_tokens_all_models": "377544.33",
+                    "total_total_tokens_all_models": "4530532.0",
+                    "avg_turn_count": "22.58",
+                    "total_turn_count": "271.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/e5442618-79a9-40f6-83b5-95497d0bd979_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb",
+        "total_rows": 42
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/e5442618-79a9-40f6-83b5-95497d0bd979_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb7d1d6e097fbf7fd7fbc7d9859f8913d2f65fe748c4f82fa6ec56254bc6adb
+size 6162708

data/frontiermath/anthropic/claude-opus-4-20250514/e5645312-da64-4834-87e2-9dcc3759a124.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786113459.500332",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786113459.500332",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.08333333333333333,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.08333333333333334
+                    },
+                    "num_samples": 12
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "11",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "503489.92",
+                    "total_total_tokens_target_model": "6041879.0",
+                    "avg_total_tokens_other_models": "107.25",
+                    "total_total_tokens_other_models": "1287.0",
+                    "avg_total_tokens_all_models": "503597.17",
+                    "total_total_tokens_all_models": "6043166.0",
+                    "avg_turn_count": "24.42",
+                    "total_turn_count": "293.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/e5645312-da64-4834-87e2-9dcc3759a124_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b",
+        "total_rows": 12
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/e5645312-da64-4834-87e2-9dcc3759a124_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b
+size 2259562

data/frontiermath/anthropic/claude-opus-4-20250514/e58b2c19-d015-4f82-b055-dc05e92facf3.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786117772.886222",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786117772.886222",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.08333333333333333,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.08333333333333334
+                    },
+                    "num_samples": 12
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "11",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "503489.92",
+                    "total_total_tokens_target_model": "6041879.0",
+                    "avg_total_tokens_other_models": "107.25",
+                    "total_total_tokens_other_models": "1287.0",
+                    "avg_total_tokens_all_models": "503597.17",
+                    "total_total_tokens_all_models": "6043166.0",
+                    "avg_turn_count": "24.42",
+                    "total_turn_count": "293.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/e58b2c19-d015-4f82-b055-dc05e92facf3_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b",
+        "total_rows": 12
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/e58b2c19-d015-4f82-b055-dc05e92facf3_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b
+size 2259562

data/frontiermath/anthropic/claude-opus-4-20250514/e8771ad5-5dbc-4551-8eb5-df2e6d124100.json

@@ -0,0 +1,107 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-20250514/1786111865.145941",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786111865.145941",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-20250514",
+        "id": "anthropic/claude-opus-4-20250514",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 6,
+                "sample_ids": [
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.08333333333333333,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.08333333333333334
+                    },
+                    "num_samples": 12
+                },
+                "details": {
+                    "total_matched_trajectories": "12",
+                    "stopping_reason_count_repetition_guard": "11",
+                    "stopping_reason_count_completed_on_successful_submit": "1",
+                    "avg_total_tokens_target_model": "503489.92",
+                    "total_total_tokens_target_model": "6041879.0",
+                    "avg_total_tokens_other_models": "107.25",
+                    "total_total_tokens_other_models": "1287.0",
+                    "avg_total_tokens_all_models": "503597.17",
+                    "total_total_tokens_all_models": "6043166.0",
+                    "avg_turn_count": "24.42",
+                    "total_turn_count": "293.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-20250514/e8771ad5-5dbc-4551-8eb5-df2e6d124100_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b",
+        "total_rows": 12
+    }
+}

data/frontiermath/anthropic/claude-opus-4-20250514/e8771ad5-5dbc-4551-8eb5-df2e6d124100_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64738dd16f2588fdfe5dd63942f851c932712f6477c95a9a0b82b376fef8d34b
+size 2259562

data/frontiermath/anthropic/claude-opus-4-6/8e1c83ac-f39a-4e04-82b3-9bf0f94b733c.json

@@ -0,0 +1,100 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786117716.442563",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786117716.442563",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+1ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 3,
+                "sample_ids": [
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 3
+                },
+                "details": {
+                    "total_matched_trajectories": "3",
+                    "stopping_reason_count_completed_on_successful_submit": "3",
+                    "avg_total_tokens_target_model": "343101.33",
+                    "total_total_tokens_target_model": "1029304.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "343101.33",
+                    "total_total_tokens_all_models": "1029304.0",
+                    "avg_turn_count": "10.00",
+                    "total_turn_count": "30.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/8e1c83ac-f39a-4e04-82b3-9bf0f94b733c_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff",
+        "total_rows": 3
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/8e1c83ac-f39a-4e04-82b3-9bf0f94b733c_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff
+size 475638

data/frontiermath/anthropic/claude-opus-4-6/a4afc074-486d-4b87-8bcb-e24f39a3e348.json

@@ -0,0 +1,99 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786113440.939973",
+    "evaluation_timestamp": "1775606918.0",
+    "retrieved_timestamp": "1786113440.939973",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWA2",
+                    "CWD31"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775606918.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 4
+                },
+                "details": {
+                    "total_matched_trajectories": "4",
+                    "stopping_reason_count_completed_on_successful_submit": "4",
+                    "avg_total_tokens_target_model": "601429.50",
+                    "total_total_tokens_target_model": "2405718.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "601429.50",
+                    "total_total_tokens_all_models": "2405718.0",
+                    "avg_turn_count": "20.75",
+                    "total_turn_count": "83.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/a4afc074-486d-4b87-8bcb-e24f39a3e348_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4",
+        "total_rows": 4
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/a4afc074-486d-4b87-8bcb-e24f39a3e348_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4
+size 648499

data/frontiermath/anthropic/claude-opus-4-6/a9d03dcf-aae3-4844-8c45-983d2f1563e1.json

@@ -0,0 +1,100 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786111834.398137",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786111834.398137",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+1ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 3,
+                "sample_ids": [
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 3
+                },
+                "details": {
+                    "total_matched_trajectories": "3",
+                    "stopping_reason_count_completed_on_successful_submit": "3",
+                    "avg_total_tokens_target_model": "343101.33",
+                    "total_total_tokens_target_model": "1029304.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "343101.33",
+                    "total_total_tokens_all_models": "1029304.0",
+                    "avg_turn_count": "10.00",
+                    "total_turn_count": "30.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/a9d03dcf-aae3-4844-8c45-983d2f1563e1_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff",
+        "total_rows": 3
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/a9d03dcf-aae3-4844-8c45-983d2f1563e1_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff
+size 475638

data/frontiermath/anthropic/claude-opus-4-6/acc3757d-100a-40e1-8f09-fb3fb5f6fbcd.json

@@ -0,0 +1,100 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786113438.136655",
+    "evaluation_timestamp": "1775604415.0",
+    "retrieved_timestamp": "1786113438.136655",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+1ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 3,
+                "sample_ids": [
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775604415.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 3
+                },
+                "details": {
+                    "total_matched_trajectories": "3",
+                    "stopping_reason_count_completed_on_successful_submit": "3",
+                    "avg_total_tokens_target_model": "343101.33",
+                    "total_total_tokens_target_model": "1029304.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "343101.33",
+                    "total_total_tokens_all_models": "1029304.0",
+                    "avg_turn_count": "10.00",
+                    "total_turn_count": "30.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/acc3757d-100a-40e1-8f09-fb3fb5f6fbcd_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff",
+        "total_rows": 3
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/acc3757d-100a-40e1-8f09-fb3fb5f6fbcd_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5409e6d26d5a0c46094259728c7c1538abca2152e7033cb5c87a4a109e6f3cff
+size 475638

data/frontiermath/anthropic/claude-opus-4-6/d49efcd2-4285-4d09-add9-ee82691010cf.json

@@ -0,0 +1,99 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786111838.009588",
+    "evaluation_timestamp": "1775606918.0",
+    "retrieved_timestamp": "1786111838.009588",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWA2",
+                    "CWD31"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775606918.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 4
+                },
+                "details": {
+                    "total_matched_trajectories": "4",
+                    "stopping_reason_count_completed_on_successful_submit": "4",
+                    "avg_total_tokens_target_model": "601429.50",
+                    "total_total_tokens_target_model": "2405718.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "601429.50",
+                    "total_total_tokens_all_models": "2405718.0",
+                    "avg_turn_count": "20.75",
+                    "total_turn_count": "83.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/d49efcd2-4285-4d09-add9-ee82691010cf_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4",
+        "total_rows": 4
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/d49efcd2-4285-4d09-add9-ee82691010cf_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4
+size 648499

data/frontiermath/anthropic/claude-opus-4-6/de31a13d-6485-4780-bde2-d20bcbb301aa.json

@@ -0,0 +1,99 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/anthropic_claude-opus-4-6/1786117719.359332",
+    "evaluation_timestamp": "1775606918.0",
+    "retrieved_timestamp": "1786117719.359332",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "anthropic/claude-opus-4-6",
+        "id": "anthropic/claude-opus-4-6",
+        "developer": "anthropic",
+        "inference_platform": "anthropic",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWA2",
+                    "CWD31"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775606918.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 4
+                },
+                "details": {
+                    "total_matched_trajectories": "4",
+                    "stopping_reason_count_completed_on_successful_submit": "4",
+                    "avg_total_tokens_target_model": "601429.50",
+                    "total_total_tokens_target_model": "2405718.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "601429.50",
+                    "total_total_tokens_all_models": "2405718.0",
+                    "avg_turn_count": "20.75",
+                    "total_turn_count": "83.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": false,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"xhigh\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/anthropic/claude-opus-4-6/de31a13d-6485-4780-bde2-d20bcbb301aa_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4",
+        "total_rows": 4
+    }
+}

data/frontiermath/anthropic/claude-opus-4-6/de31a13d-6485-4780-bde2-d20bcbb301aa_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5741f2ddca66474718ab75bd731d52ba624929e9868874603ad98267350ab8e4
+size 648499

data/frontiermath/openai/gpt-5-2025-08-07/25b3b8b0-e40f-42d0-895e-89aff557ca79.json

@@ -0,0 +1,116 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5-2025-08-07/1786111804.545191",
+    "evaluation_timestamp": "1775078458.0",
+    "retrieved_timestamp": "1786111804.545191",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5-2025-08-07",
+        "id": "openai/gpt-5-2025-08-07",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 12,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3",
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775078458.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.6666666666666666,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.13725299209264563
+                    },
+                    "num_samples": 120
+                },
+                "details": {
+                    "total_matched_trajectories": "60",
+                    "stopping_reason_count_repetition_guard": "20",
+                    "stopping_reason_count_completed_on_successful_submit": "40",
+                    "avg_total_tokens_target_model": "778666.62",
+                    "total_total_tokens_target_model": "46719997.0",
+                    "avg_total_tokens_other_models": "39.00",
+                    "total_total_tokens_other_models": "2340.0",
+                    "avg_total_tokens_all_models": "778705.62",
+                    "total_total_tokens_all_models": "46722337.0",
+                    "avg_turn_count": "20.47",
+                    "total_turn_count": "1228.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5-2025-08-07/25b3b8b0-e40f-42d0-895e-89aff557ca79_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc",
+        "total_rows": 120
+    }
+}

data/frontiermath/openai/gpt-5-2025-08-07/25b3b8b0-e40f-42d0-895e-89aff557ca79_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc
+size 34029287

data/frontiermath/openai/gpt-5-2025-08-07/3a600726-32cb-4b1f-8dd7-26f6c60fda24.json

@@ -0,0 +1,116 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5-2025-08-07/1786113417.676557",
+    "evaluation_timestamp": "1775078458.0",
+    "retrieved_timestamp": "1786113417.676557",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5-2025-08-07",
+        "id": "openai/gpt-5-2025-08-07",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 12,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3",
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775078458.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.6666666666666666,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.13725299209264563
+                    },
+                    "num_samples": 120
+                },
+                "details": {
+                    "total_matched_trajectories": "60",
+                    "stopping_reason_count_repetition_guard": "20",
+                    "stopping_reason_count_completed_on_successful_submit": "40",
+                    "avg_total_tokens_target_model": "778666.62",
+                    "total_total_tokens_target_model": "46719997.0",
+                    "avg_total_tokens_other_models": "39.00",
+                    "total_total_tokens_other_models": "2340.0",
+                    "avg_total_tokens_all_models": "778705.62",
+                    "total_total_tokens_all_models": "46722337.0",
+                    "avg_turn_count": "20.47",
+                    "total_turn_count": "1228.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5-2025-08-07/3a600726-32cb-4b1f-8dd7-26f6c60fda24_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc",
+        "total_rows": 120
+    }
+}

data/frontiermath/openai/gpt-5-2025-08-07/3a600726-32cb-4b1f-8dd7-26f6c60fda24_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc
+size 34029287

data/frontiermath/openai/gpt-5-2025-08-07/9386221b-7497-4dfc-ab5e-5ba7658105a6.json

@@ -0,0 +1,116 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5-2025-08-07/1786117686.866183",
+    "evaluation_timestamp": "1775078458.0",
+    "retrieved_timestamp": "1786117686.866183",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.202,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5-2025-08-07",
+        "id": "openai/gpt-5-2025-08-07",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 12,
+                "sample_ids": [
+                    "ALL3",
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1",
+                    "ONO3",
+                    "OVE2",
+                    "PLD1",
+                    "PTA1",
+                    "RAP1",
+                    "RSG1",
+                    "TIK2"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1775078458.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 0.6666666666666666,
+                "uncertainty": {
+                    "standard_error": {
+                        "value": 0.13725299209264563
+                    },
+                    "num_samples": 120
+                },
+                "details": {
+                    "total_matched_trajectories": "60",
+                    "stopping_reason_count_repetition_guard": "20",
+                    "stopping_reason_count_completed_on_successful_submit": "40",
+                    "avg_total_tokens_target_model": "778666.62",
+                    "total_total_tokens_target_model": "46719997.0",
+                    "avg_total_tokens_other_models": "39.00",
+                    "total_total_tokens_other_models": "2340.0",
+                    "avg_total_tokens_all_models": "778705.62",
+                    "total_total_tokens_all_models": "46722337.0",
+                    "avg_turn_count": "20.47",
+                    "total_turn_count": "1228.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5-2025-08-07/9386221b-7497-4dfc-ab5e-5ba7658105a6_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc",
+        "total_rows": 120
+    }
+}

data/frontiermath/openai/gpt-5-2025-08-07/9386221b-7497-4dfc-ab5e-5ba7658105a6_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e377cf148228f271d458b2676c1a72f1299c2a4935d2ed863b7895c453b3dfc
+size 34029287

data/frontiermath/openai/gpt-5.4-2026-03-05/0d120bcc-ed75-4c6f-bef6-74186cfd277a.json

@@ -0,0 +1,102 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5.4-2026-03-05/1786117586.200713",
+    "evaluation_timestamp": "1776165072.0",
+    "retrieved_timestamp": "1786117586.200713",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.206.dev19+g6beb95d9,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5.4-2026-03-05",
+        "id": "openai/gpt-5.4-2026-03-05",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+1ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWA2",
+                    "LTI1"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1776165072.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 2
+                },
+                "details": {
+                    "total_matched_trajectories": "2",
+                    "stopping_reason_count_completed_on_successful_submit": "2",
+                    "avg_total_tokens_target_model": "59181.00",
+                    "total_total_tokens_target_model": "118362.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "59181.00",
+                    "total_total_tokens_all_models": "118362.0",
+                    "avg_turn_count": "6.50",
+                    "total_turn_count": "13.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5.4-2026-03-05/0d120bcc-ed75-4c6f-bef6-74186cfd277a_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "bdbdb499168956ef7f82ad3d5fe2ed50109f5c4b32cb1ce2a4277fa1f1615651",
+        "total_rows": 2
+    }
+}

data/frontiermath/openai/gpt-5.4-2026-03-05/0d120bcc-ed75-4c6f-bef6-74186cfd277a_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdbdb499168956ef7f82ad3d5fe2ed50109f5c4b32cb1ce2a4277fa1f1615651
+size 151388

data/frontiermath/openai/gpt-5.4-2026-03-05/2d5ce0c4-d254-42ef-aa30-4cae9bc2f828.json

@@ -0,0 +1,104 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5.4-2026-03-05/1786117606.714761",
+    "evaluation_timestamp": "1776258608.0",
+    "retrieved_timestamp": "1786117606.714761",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.206.dev19+g6beb95d9,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5.4-2026-03-05",
+        "id": "openai/gpt-5.4-2026-03-05",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+4ep/a7843388 for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 4,
+                "sample_ids": [
+                    "CWA2",
+                    "CWD31",
+                    "FMT1",
+                    "LTI1"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1776258608.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 16
+                },
+                "details": {
+                    "total_matched_trajectories": "8",
+                    "stopping_reason_count_completed_on_successful_submit": "8",
+                    "avg_total_tokens_target_model": "45446.75",
+                    "total_total_tokens_target_model": "363574.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "45446.75",
+                    "total_total_tokens_all_models": "363574.0",
+                    "avg_turn_count": "6.25",
+                    "total_turn_count": "50.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5.4-2026-03-05/2d5ce0c4-d254-42ef-aa30-4cae9bc2f828_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "b9eb3b24467ae8d559a40ff2f1f7c2db852f78c4f6dda7dc54d9d5d816121321",
+        "total_rows": 16
+    }
+}

data/frontiermath/openai/gpt-5.4-2026-03-05/2d5ce0c4-d254-42ef-aa30-4cae9bc2f828_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9eb3b24467ae8d559a40ff2f1f7c2db852f78c4f6dda7dc54d9d5d816121321
+size 1524208

data/frontiermath/openai/gpt-5.4-2026-03-05/9337ec6c-8c7c-4b77-8f89-6cc789b81dae.json

@@ -0,0 +1,102 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5.4-2026-03-05/1786117592.653393",
+    "evaluation_timestamp": "1776165072.0",
+    "retrieved_timestamp": "1786117592.653393",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.206.dev19+g6beb95d9,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5.4-2026-03-05",
+        "id": "openai/gpt-5.4-2026-03-05",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+2ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWD31",
+                    "FMT1"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1776165072.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 4
+                },
+                "details": {
+                    "total_matched_trajectories": "4",
+                    "stopping_reason_count_completed_on_successful_submit": "4",
+                    "avg_total_tokens_target_model": "154780.25",
+                    "total_total_tokens_target_model": "619121.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "154780.25",
+                    "total_total_tokens_all_models": "619121.0",
+                    "avg_turn_count": "12.50",
+                    "total_turn_count": "50.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5.4-2026-03-05/9337ec6c-8c7c-4b77-8f89-6cc789b81dae_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "ae6ff23b61fc286f033378d8de23607c291d12d3760e99b6ce59cabe814895e0",
+        "total_rows": 4
+    }
+}

data/frontiermath/openai/gpt-5.4-2026-03-05/9337ec6c-8c7c-4b77-8f89-6cc789b81dae_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae6ff23b61fc286f033378d8de23607c291d12d3760e99b6ce59cabe814895e0
+size 498651

data/frontiermath/openai/gpt-5.4-2026-03-05/b284273f-18c2-46ec-93eb-b14a942981a4.json

@@ -0,0 +1,90 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5.4-2026-03-05/1786117609.961544",
+    "evaluation_timestamp": "1776256372.0",
+    "retrieved_timestamp": "1786117609.961544",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.206.dev19+g6beb95d9,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5.4-2026-03-05",
+        "id": "openai/gpt-5.4-2026-03-05",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C/+5ep for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 2,
+                "sample_ids": [
+                    "CWA2",
+                    "LTI1"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1776256372.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 10
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5.4-2026-03-05/b284273f-18c2-46ec-93eb-b14a942981a4_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "52a4c11caf6d4b354be572c60360b353dc63a3550200fad66d0f7341f577532e",
+        "total_rows": 10
+    }
+}

data/frontiermath/openai/gpt-5.4-2026-03-05/b284273f-18c2-46ec-93eb-b14a942981a4_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52a4c11caf6d4b354be572c60360b353dc63a3550200fad66d0f7341f577532e
+size 726925

data/frontiermath/openai/gpt-5.4-2026-03-05/c57a5f4e-9b90-40eb-9c08-bf9057db3a2e.json

@@ -0,0 +1,101 @@
+{
+    "schema_version": "0.3.0",
+    "evaluation_id": "frontiermath/openai_gpt-5.4-2026-03-05/1786117584.148378",
+    "evaluation_timestamp": "1776258608.0",
+    "retrieved_timestamp": "1786117584.148378",
+    "source_metadata": {
+        "source_name": "inspect_ai",
+        "source_type": "evaluation_run",
+        "source_organization_name": "unknown",
+        "evaluator_relationship": "third_party"
+    },
+    "eval_library": {
+        "name": "inspect_ai",
+        "version": "inspect_ai:0.3.206.dev19+g6beb95d9,inference_scaling_paper:0.1.0"
+    },
+    "model_info": {
+        "name": "openai/gpt-5.4-2026-03-05",
+        "id": "openai/gpt-5.4-2026-03-05",
+        "developer": "openai",
+        "inference_platform": "openai",
+        "additional_details": {
+            "deployment_type": "externally_managed",
+            "model_availability": "closed_weights"
+        }
+    },
+    "evaluation_results": [
+        {
+            "evaluation_name": "accuracy on frontiermath/S-adaptive+C for scorer verification_code",
+            "source_data": {
+                "dataset_name": "frontiermath",
+                "source_type": "hf_dataset",
+                "samples_number": 1,
+                "sample_ids": [
+                    "ONO3"
+                ],
+                "additional_details": {
+                    "shuffled": "False"
+                }
+            },
+            "evaluation_timestamp": "1776258608.0",
+            "metric_config": {
+                "evaluation_description": "accuracy",
+                "lower_is_better": false,
+                "score_type": "continuous",
+                "min_score": 0.0,
+                "max_score": 1.0
+            },
+            "score_details": {
+                "score": 1.0,
+                "uncertainty": {
+                    "num_samples": 10
+                },
+                "details": {
+                    "total_matched_trajectories": "5",
+                    "stopping_reason_count_completed_on_successful_submit": "5",
+                    "avg_total_tokens_target_model": "561340.80",
+                    "total_total_tokens_target_model": "2806704.0",
+                    "avg_total_tokens_other_models": "0.00",
+                    "total_total_tokens_other_models": "0.0",
+                    "avg_total_tokens_all_models": "561340.80",
+                    "total_total_tokens_all_models": "2806704.0",
+                    "avg_turn_count": "27.20",
+                    "total_turn_count": "136.0"
+                }
+            },
+            "generation_config": {
+                "generation_args": {
+                    "reasoning": true,
+                    "agentic_eval_config": {
+                        "available_tools": []
+                    },
+                    "eval_plan": {
+                        "name": "plan",
+                        "steps": [
+                            "{\"solver\": \"react\", \"params\": {\"name\": null, \"description\": null, \"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"model\": null, \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"retry_refusals\": null, \"compaction\": \"CompactionSummary\", \"truncation\": \"disabled\", \"approval\": null}, \"params_passed\": {\"prompt\": [\"\", \"\\nYou are part of a multi-agent system designed to make agent coordination and execution easy. Agents uses two primary abstraction: **Agents** and **Handoffs**. An agent encompasses instructions and tools and can hand off a conversation to another agent when appropriate. Handoffs are achieved by calling a handoff function,generally named `transfer_to_<agent_name>`. Transfers between agents are handled seamlessly in the background; do not mention or draw attention to these transfers in your conversation with the user.\\n\", \"\\nYou are a helpful assistant attempting to submit the best possible answer. You have several tools available to help with finding the answer. You will see the result of tool calls right after sending the message. If you need to perform multiple actions, you can always send more messages with additional tool calls. Do some reasoning before your actions, describing what tool calls you are going to use and how they fit into your plan.\\n\", \"\\nWhen you have completed the task and have an answer, call the {submit}() tool to report it.\\n\"], \"tools\": [{\"type\": \"tool\", \"name\": \"python\", \"params\": {\"timeout\": 180}}], \"attempts\": [999, \"wrapper\", \"to_float\"], \"submit\": [null, null, {\"type\": \"tool\", \"name\": \"inference_scaling_paper/submit_answer\", \"params\": {}}, false, \"\\n\\n\", false], \"on_continue\": \"_guard\", \"compaction\": \"CompactionSummary\"}}"
+                        ],
+                        "config": {
+                            "attempt_timeout": "1200",
+                            "reasoning_effort": "\"high\"",
+                            "reasoning_tokens": "64000"
+                        }
+                    },
+                    "eval_limits": {
+                        "token_limit": 10000000
+                    },
+                    "sandbox": {}
+                },
+                "additional_details": {
+                    "reasoning_effort": "\"high\""
+                }
+            }
+        }
+    ],
+    "detailed_evaluation_results": {
+        "format": "jsonl",
+        "file_path": "data/frontiermath/openai/gpt-5.4-2026-03-05/c57a5f4e-9b90-40eb-9c08-bf9057db3a2e_samples.jsonl",
+        "hash_algorithm": "sha256",
+        "checksum": "b6bc8967af581b6970e0f9fb840c937bf16e284ef60d8e7e9d3b5b51d7789075",
+        "total_rows": 10
+    }
+}

data/frontiermath/openai/gpt-5.4-2026-03-05/c57a5f4e-9b90-40eb-9c08-bf9057db3a2e_samples.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6bc8967af581b6970e0f9fb840c937bf16e284ef60d8e7e9d3b5b51d7789075
+size 2072553