Dataset Viewer
Auto-converted to Parquet Duplicate
paper_id
stringlengths
13
13
title
stringlengths
19
145
paper_url
stringlengths
32
32
authors
listlengths
1
40
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
671
1.92k
keywords
listlengths
0
0
TL;DR
large_stringclasses
0 values
submission_number
int64
1
856
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
1 value
colm2026-0001
FlexRouter: Learning Complementary Model Sets for Flexible LLM Routing
null
[ "Wang Wei", "Hongzheng Yang", "Tiankai Yang", "Samyadeep Basu", "Hongjie Chen", "Yue Zhao", "Franck Dernoncourt", "Ryan A. Rossi", "Hoda Eldardiry" ]
null
null
null
[]
null
1
null
null
colm2026-0002
CT-$\Delta$Bench: A Benchmark for Longitudinal 3D Medical Imaging Difference Reporting with Vision-Language Models
null
[ "Kegeng Tang", "Jingbo Wang", "Shaogang Ren", "Zihao WANG" ]
null
null
null
[]
null
2
null
null
colm2026-0003
No, there are no grammatical ``grandmother neurons'' in LLMs
null
[ "Linyang He", "Nima Mesgarani" ]
null
null
null
[]
null
3
null
null
colm2026-0004
Beyond Logit Adjustment: A Residual Decomposition Framework for Long-Tailed Reranking
https://arxiv.org/abs/2604.01506
[ "Zhanliang Wang", "Hongzhuo Chen", "Quan Minh Nguyen", "Mian Umair Ahsan", "Kai Wang" ]
null
null
Long-tailed classification, where a small number of frequent classes dominate many rare ones, remains challenging because models systematically favor frequent classes at inference time. Existing post-hoc methods such as logit adjustment address this by adding a fixed classwise offset to the base-model logits. However, ...
[]
null
4
2604.01506
title_snapshot
colm2026-0005
Do LLMs Recognize Your Latent Preferences? A Benchmark for Latent Information Discovery in Personalized Interaction
https://arxiv.org/abs/2510.17132
[ "Ioannis Tsaknakis", "Bingqing Song", "Shuyu Gan", "Jiangweizhi Peng", "Dongyeop Kang", "Alfredo Garcia", "Gaowen Liu", "Charles Fleming", "Mingyi Hong" ]
null
null
Large Language Models (LLMs) excel at producing broadly relevant text, but this generality becomes a limitation when user-specific preferences are required, such as recommending restaurants or planning travel. In these scenarios, users rarely articulate every preference explicitly; instead, much of what they care about...
[]
null
5
2510.17132
title_snapshot
colm2026-0006
BibTeX Citation Hallucinations in Scientific Publishing Agents: Evaluation and Mitigation
https://arxiv.org/abs/2604.03159
[ "Delip Rao", "Chris Callison-Burch" ]
null
null
Large language models with web search are increasingly used in scientific publishing agents, yet they still produce BibTeX entries with pervasive field-level errors. Prior evaluations tested base models without search, which does not reflect current practice. We construct a benchmark of 931 papers across four scientifi...
[]
null
6
2604.03159
title_snapshot
colm2026-0007
KronQ: LLM Quantization via Kronecker-Factored Hessian
https://arxiv.org/abs/2607.07964
[ "Donghyun Lee", "Yuhang Li", "Ruokai Yin", "Priyadarshini Panda" ]
null
null
Post-training quantization (PTQ) is a widely adopted technique for compressing large language models (LLMs) without retraining. Existing second-order PTQ methods, including GPTQ, construct quantization objectives exclusively from input activation statistics, effectively assuming that all output channels contribute equa...
[]
null
7
2607.07964
title_snapshot
colm2026-0008
The Tool Illusion: Rethinking Tool Use in Web Agents
https://arxiv.org/abs/2604.03465
[ "Renze Lou", "Baolin Peng", "Wenlin Yao", "Qianhui Wu", "Hao Cheng", "Suman Nath", "Wenpeng Yin", "Jianfeng Gao" ]
null
null
As web agents rapidly evolve, an increasing body of work has moved beyond conventional atomic browser interactions and explored tool use as a higher-level action paradigm. Although prior studies have shown the promise of tools, their conclusions are often drawn from limited experimental scales and sometimes non-compara...
[]
null
8
2604.03465
title_snapshot
colm2026-0009
PreMoE: Proactive Inference for Efficient Mixture-of-Experts
https://arxiv.org/abs/2505.17639
[ "Zehua Pei", "Ying Zhang", "Hui-Ling Zhen", "Tao Yuan", "Xianzhi Yu", "Zhenhua Dong", "Sinno Jialin Pan", "Mingxuan Yuan", "Bei Yu" ]
null
null
Mixture-of-Experts (MoE) models offer dynamic computation, but are typically deployed as static full-capacity models, missing opportunities for deployment-specific specialization. We introduce PreMoE, a training-free framework that proactively compiles sparse MoE variants for targeted deployment scenarios. At its core ...
[]
null
9
2505.17639
title_snapshot
colm2026-0010
Delay, Plateau, or Collapse: Evaluating the Impact of Systematic Verification Error on RLVR
https://arxiv.org/abs/2605.02909
[ "Kazuki Egashira", "Mark Vero", "Jasper Dekoninck", "Florian E. Dorner", "Robin Staab", "Martin Vechev" ]
null
null
Reinforcement Learning with Verifiable Rewards (RLVR) has become a powerful approach for improving the reasoning capabilities of large language models (LLMs). While RLVR is designed for tasks with verifiable ground-truth answers, real-world verifiers (e.g., static code checkers) can introduce errors into the reward sig...
[]
null
10
2605.02909
title_snapshot
colm2026-0011
It’s How You Ask: Gendered Linguistic Bias in LLMs
null
[ "Katherine Van Koevering", "Anjalie Field" ]
null
null
null
[]
null
11
null
null
colm2026-0012
ConsumerBench: Benchmarking Generative AI Applications on End-User Devices
https://arxiv.org/abs/2506.17538
[ "Yile Gu", "Rohan Kadekodi", "Sahi Chitrapu", "Aradhya Agrawal", "Camille Sawa", "Hoang Doan Nguyen", "Keisuke Kamahori", "Yiyu Liu", "Baris Kasikci" ]
null
null
The recent shift in Generative AI (GenAI) applications from cloud-only environments to end-user devices introduces new challenges in resource management, system efficiency, and user experience. This paper presents ConsumerBench, a comprehensive benchmarking framework designed to evaluate the system efficiency and respo...
[]
null
12
2506.17538
title_snapshot
colm2026-0013
RefusalGuard: Geometry-Preserving Fine-Tuning for Safety in LLMs
https://arxiv.org/abs/2605.01913
[ "Sadia Asif", "Mohammad Mohammadi Amiri" ]
null
null
Fine-tuning safety-aligned language models for downstream tasks often leads to substantial degradation of refusal behavior, making models vulnerable to adversarial misuse. While prior work has shown that safety-relevant features are encoded in structured representations within the model's activation space, how these re...
[]
null
13
2605.01913
title_snapshot
colm2026-0014
Filtered Reasoning Score: Evaluating Reasoning Quality on a Model's Most-Confident Traces
https://arxiv.org/abs/2604.11996
[ "Manas Pathak", "Xingyao Chen", "Shuozhe Li", "Amy Zhang", "Liu Leqi" ]
null
null
Should we trust Large Language Models (LLMs) with high accuracy? LLMs achieve high accuracy on reasoning benchmarks, but correctness alone does not reveal the quality of the reasoning used to produce it. This highlights a fundamental limitation of outcome-based evaluation: models may arrive at correct answers through f...
[]
null
14
2604.11996
title_snapshot
colm2026-0015
OpenStamp: A Watermark for Open-Source Language Models
null
[ "Miroojin Bakshi", "Saksham Rastogi", "Danish Pruthi" ]
null
null
null
[]
null
15
null
null
colm2026-0016
In-Context Examples Suppress Scientific Knowledge Recall in LLMs
https://arxiv.org/abs/2604.27540
[ "Chaemin Jang", "Woojin Park", "Hyeok Yun", "Dongman Lee", "Jihee Kim" ]
null
null
Scientific reasoning rarely stops at what is directly observable; it often requires uncovering hidden structure from data. From estimating reaction constants in chemistry to inferring demand elasticities in economics, this latent structure recovery is what distinguishes scientific reasoning from curve fitting. Large la...
[]
null
16
2604.27540
title_snapshot
colm2026-0017
PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost
https://arxiv.org/abs/2603.21383
[ "Junkeun Yi", "Damon Mosk-Aoyama", "Baihe Huang", "Ritu Gala", "Charles Wang", "Sugam Devare", "Khushi Bhardwaj", "Abhibha Gupta", "Oleksii Kuchaiev", "Jiantao Jiao", "Jian Zhang", "Venkat Krishna Srinivasan" ]
null
null
Post-training for long-horizon agentic tasks has a tension between compute efficiency and generalization. While supervised fine-tuning (SFT) is compute efficient, it often suffers from out-of-domain (OOD) degradation. Conversely, end-to-end reinforcement learning (E2E RL) preserves OOD capabilities, but incurs high com...
[]
null
17
2603.21383
title_snapshot
colm2026-0018
Legibility is Not Interpretability: Evaluating Judged Importance versus Actual Importance in Chain-Of-Thought Reasoning Steps
null
[ "Kevin Du", "Alexander Miserlis Hoyle", "Laura Ruis", "Acyr Locatelli" ]
null
null
null
[]
null
18
null
null
colm2026-0019
Learning Steerable Clarification Policies with Collaborative Self-play
https://arxiv.org/abs/2512.04068
[ "Jonathan Berant", "Maximillian Chen", "Adam Fisch", "Reza Aghajani", "Fantine Eri Huot", "Mirella Lapata", "Jacob Eisenstein" ]
null
null
To handle underspecified or ambiguous queries, AI assistants need a policy for managing their uncertainty to determine (a) when to guess the user intent and answer directly, (b) when to enumerate and answer multiple possible intents, and (c) when to ask a clarifying question. However, such policies are contextually dep...
[]
null
19
2512.04068
title_snapshot
colm2026-0020
Process-Oriented Evaluation of AI-Assisted Scientific Writing
https://arxiv.org/abs/2606.15583
[ "Patrick Queiroz Da Silva", "Sanchaita Hazra", "Doeun Lee", "Sachin Kumar", "Bodhisattwa Prasad Majumder" ]
null
null
Bad writing hinders the publication of science. The role of artificial intelligence (AI) in generating and editing scientific texts remains unsettled. Abstracts serve as the critical gateway to scientific manuscripts, often shaping readers' interest. We inspect how individuals revise AI-generated abstracts compared to ...
[]
null
20
2606.15583
title_snapshot
colm2026-0021
Visual Aesthetic Benchmark: Can Frontier Models Judge Beauty?
https://arxiv.org/abs/2605.12684
[ "Yichen Feng", "Yuetai Li", "Chunjiang Liu", "Fengqing Jiang", "Yue Huang", "Yuanyuan Chen", "Hang Hua", "Zhengqing Yuan", "Kaiyuan Zheng", "Luyao Niu", "Bhaskar Ramasubramanian", "Basel Alomair", "Xiangliang Zhang", "Misha Sra", "Zichen Chen", "Radha Poovendran", "Zhangchen Xu" ]
null
null
Multimodal large language models (MLLMs) are now routinely deployed for visual understanding, generation, and curation. A substantial fraction of these applications require an explicit aesthetic judgment. Most existing solutions reduce this judgment to predicting a scalar score for a single image. We first ask whether ...
[]
null
21
2605.12684
title_snapshot
colm2026-0022
Clarify or Answer: Reinforcement Learning for Agentic VQA with Context Under-specification
https://arxiv.org/abs/2601.16400
[ "Zongwan Cao", "Bingbing Wen", "Lucy Lu Wang" ]
null
null
Real-world visual question answering (VQA) is often context-dependent: an image-question pair may be under-specified, such that the correct answer depends on external information that is not observable in the image. In such cases, directly answering can lead to confident but incorrect predictions. We propose CoA(Clarif...
[]
null
22
2601.16400
title_snapshot
colm2026-0023
MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs
https://arxiv.org/abs/2605.07305
[ "Hsin-Ling Hsu", "Zizheng Wang", "Donghua Zhang", "Nai-Chia Chen", "Jerry Wang", "Jun-En Ding", "Chia-Hsuan Hsu", "Guoan Wang", "Feng Liu", "Fang Ming Hung", "Chenwei Wu", "Liyue Shen" ]
null
null
Most existing LLM diagnoses are evaluated on static, single-turn settings where complete patient information is provided upfront, an oversimplification of real clinical practice. We study active diagnosis: the real-life clinical process of starting from initial observation, ordering tests, interpreting results, and upd...
[]
null
23
2605.07305
title_snapshot
colm2026-0024
When Verification Fails: How Compositionally Infeasible Claims Escape Rejection
https://arxiv.org/abs/2604.10990
[ "Muxin Liu", "Delip Rao", "Grace Kim", "Chris Callison-Burch" ]
null
null
Scientific claim verification, the task of determining whether claims are entailed by scientific evidence, is fundamental to establishing discoveries in evidence while preventing misinformation. This process involves evaluating each asserted constraint against validated evidence. Under the Closed-World Assumption (CWA)...
[]
null
24
2604.10990
title_snapshot
colm2026-0025
Emergent Unfaithfulness: How Alignment Training Causes Language Models to Silently Override Task Faithfulness
null
[ "Pardis Sadat Zahraei", "Janvijay Singh", "Gokhan Tur", "Dilek Hakkani-Tür" ]
null
null
null
[]
null
25
null
null
colm2026-0026
SimulRAG: Simulator-based RAG for Grounding LLMs in Long-form Scientific QA
https://arxiv.org/abs/2509.25459
[ "Haozhou Xu", "Dongxia Wu", "Matteo Chinazzi", "Ruijia Niu", "Rose Yu", "Yian Ma" ]
null
null
Large language models (LLMs) show promise in solving scientific problems. They can help generate long-form answers for scientific questions, which are crucial for comprehensive understanding of complex phenomena that require detailed explanations spanning multiple interconnected concepts and evidence. However, LLMs oft...
[]
null
26
2509.25459
title_snapshot
colm2026-0027
SAKE: Structured Agentic Knowledge Extrapolation for Complex LLM Reasoning via Reinforcement Learning
https://arxiv.org/abs/2505.15062
[ "Jiashu He", "Jinxuan Fan", "Bowen Jiang", "Ignacio Hounie", "Dan Roth", "Alejandro Ribeiro" ]
null
null
Knowledge extrapolation is the process of inferring novel information by combining and extending existing knowledge that is explicitly available. It is essential for solving complex questions in specialized domains where retrieving comprehensive external knowledge is impractical. We propose SAKE (Structured Agentic Kno...
[]
null
27
2505.15062
title_snapshot
colm2026-0028
FOCUS: Closed-Loop Attention Feedback for Efficient Vision-Language Understanding
null
[ "Akul Santhosh", "Aastha Jhunjhunwala" ]
null
null
null
[]
null
28
null
null
colm2026-0029
Estimating near-verbatim extraction risk in language models with decoding-constrained beam search
https://arxiv.org/abs/2603.24917
[ "A. Feder Cooper", "Mark Lemley", "Christopher De Sa", "Lea Duesterwald", "Allison Casasola", "Jamie Hayes", "Katherine Lee", "Daniel E. Ho", "Percy Liang" ]
null
null
Recent work shows that standard greedy-decoding extraction methods for quantifying memorization in LLMs miss how extraction risk varies across sequences. Probabilistic extraction -- computing the probability of generating a target suffix given a prefix under a decoding scheme -- addresses this, but is tractable only fo...
[]
null
29
2603.24917
title_snapshot
colm2026-0030
Subspace Control: Turning Constrained Model Steering into Controllable Spectral Optimization
https://arxiv.org/abs/2604.04231
[ "Yancheng Huang", "Changsheng Wang", "Chongyu Fan", "Yicheng Lang", "Bingqi Shang", "Yang Zhang", "Mingyi Hong", "Qing Qu", "Alvaro Velasquez", "Sijia Liu" ]
null
null
Foundation models, such as large language models (LLMs), are powerful but often require customization before deployment to satisfy practical constraints such as safety, privacy, and task-specific requirements, leading to "constrained" optimization problems for model steering and adaptation. However, solving such proble...
[]
null
30
2604.04231
title_snapshot
colm2026-0031
Breaking Memorization Barriers in LLM Code Fine-Tuning via Information Bottleneck for Improved Generalization
https://arxiv.org/abs/2510.16022
[ "Changsheng Wang", "Xin Chen", "Sijia Liu", "Ke Ding" ]
null
null
Adapting pretrained large language models (LLMs) to code domains via supervised fine-tuning (FT) has been commonly used for code generation. However, we identify a previously underappreciated failure mode, the memorization barrier, where strong memorization of downstream code data in the base model could trap optimizat...
[]
null
31
2510.16022
title_snapshot
colm2026-0032
Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversations
https://arxiv.org/abs/2604.27093
[ "Mingqian Zheng", "Malia Morgan", "Liwei Jiang", "Carolyn Rose", "Maarten Sap" ]
null
null
Current LLM safety alignment techniques improve model robustness against adversarial attacks, but overlook whether and how LLMs can recover helpfulness when benign users clarify their intent. We introduce CarryOnBench, the first interactive benchmark that measures whether LLMs can revise their interpretation of user in...
[]
null
32
2604.27093
title_snapshot
colm2026-0033
The Percept-V Challenge: Can Multimodal LLMs Crack Simple Perception Problems?
https://arxiv.org/abs/2508.21143
[ "Samrajnee Ghosh", "Naman Agarwal", "Ashish Goswami", "Hemanshu Garg", "Chinmay Mittal", "Parag Singla", "Mausam" ]
null
null
Cognitive science research treats visual perception, the ability to understand and make sense of a visual input, as one of the early developmental signs of intelligence. Its TVPS-4 framework categorizes and tests human perception into seven skills such as visual discrimination, and form constancy. Do Multimodal Large L...
[]
null
33
2508.21143
title_snapshot
colm2026-0034
OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Data
https://arxiv.org/abs/2603.15594
[ "Yuwen Du", "Rui Ye", "Shuo Tang", "Xinyu Zhu", "Yijun Lu", "Yuzhu Cai", "Siheng Chen" ]
null
null
Deep search capabilities have become an indispensable competency for frontier Large Language Model (LLM) agents, yet the development of high-performance search agents remains dominated by industrial giants due to a lack of transparent, high-quality training data. This persistent data scarcity has fundamentally hindered...
[]
null
34
2603.15594
title_snapshot
colm2026-0035
Weightless Fine-Tuning: Personalizing LLMs via Logit-Space Transport
null
[ "Bohan Zhang", "Anqi Ni", "Yixin Wang", "Paramveer Dhillon" ]
null
null
null
[]
null
35
null
null
colm2026-0036
Joint Optimization of Reasoning and Dual-Memory for Self-Learning Diagnostic Agent
https://arxiv.org/abs/2604.07269
[ "Bingxuan Li", "Simo Du", "Yue Guo" ]
null
null
Clinical expertise improves not only by acquiring medical knowledge, but by accumulating experience that yields reusable diagnostic patterns. Recent LLMs-based diagnostic agents have shown promising progress in clinical reasoning for decision support. However, most approaches treat cases independently, limiting experie...
[]
null
36
2604.07269
title_snapshot
colm2026-0037
MoRFI: Monotonic Sparse Autoencoder Feature Identification
https://arxiv.org/abs/2604.26866
[ "Dimitris Dimakopoulos", "Shay B. Cohen", "Ioannis Konstas" ]
null
null
Large language models (LLMs) acquire most of their factual knowledge during the pre-training stage, through next token prediction. Subsequent stages of post-training often introduce new facts outwith the parametric knowledge, giving rise to hallucinations. While it has been demonstrated that supervised fine-tuning (SFT...
[]
null
37
2604.26866
title_snapshot
colm2026-0038
English is Not All You Need: Systematically Exploring the Role of Multilinguality in LLM Post-Training
https://arxiv.org/abs/2604.13286
[ "Mehak Preet Dhaliwal", "Shashwat Ranjan Chaurasia", "Yao Qin", "Dezhi Hong", "Thomas Butler" ]
null
null
Despite the widespread multilingual deployment of large language models, post-training pipelines remain predominantly English-centric, contributing to performance disparities across languages. We present a systematic, controlled study of the interplay between training language coverage, model scale, and task domain, ba...
[]
null
38
2604.13286
title_snapshot
colm2026-0039
Scaling Self-Play with Self-Guidance
https://arxiv.org/abs/2604.20209
[ "Luke Bailey", "Kaiyue Wen", "Kefan Dong", "Tatsunori Hashimoto", "Tengyu Ma" ]
null
null
LLM self-play algorithms are notable in that, in principle, nothing bounds their learning: a Conjecturer model creates problems for a Solver, and both improve together. However, in practice, existing LLM self-play methods do not scale well with large amounts of compute, instead hitting learning plateaus. We argue this ...
[]
null
39
2604.20209
title_snapshot
colm2026-0040
Procedural Knowledge at Scale Improves Reasoning
https://arxiv.org/abs/2604.01348
[ "Di Wu", "Devendra Singh Sachan", "Wen-tau Yih", "Mingda Chen" ]
null
null
Test-time scaling has emerged as an effective way to improve language models on challenging reasoning tasks. However, most existing methods treat each problem in isolation and do not systematically reuse knowledge from prior reasoning trajectories. In particular, they underutilize procedural knowledge: how to reframe a...
[]
null
40
2604.01348
title_snapshot
colm2026-0041
Variational Co-Evolution via Reinforcement Learning
null
[ "Qiahao Zheng", "Hancong Jin", "Junbo Niu", "Yinjie Wang", "Wentao Zhang", "Ruixun Zhang" ]
null
null
null
[]
null
41
null
null
colm2026-0042
Toward Scalable Terminal Task Synthesis via Skill Graphs
https://arxiv.org/abs/2604.25727
[ "Zhiyuan Fan", "TingHao YU", "Yuanjun Cai", "JiangTaoGuan", "Yun Yang", "Dingxin Hu", "Xing W", "Zhuo Han", "feng zhang", "Lilin Wang" ]
null
null
Terminal agents have demonstrated strong potential for autonomous command-line execution, yet their training remains constrained by the scarcity of high-quality and diverse execution trajectories. Existing approaches mitigate this bottleneck by synthesizing large-scale terminal task instances for trajectory sampling. H...
[]
null
42
2604.25727
title_snapshot
colm2026-0043
How Transformers Learn to Plan via Multi-Token Prediction
https://arxiv.org/abs/2604.11912
[ "Jianhao Huang", "Zhanpeng Zhou", "Renqiu Xia", "Baharan Mirzasoleiman", "Weijie J Su", "Wei Huang" ]
null
null
While next-token prediction (NTP) has been the standard objective for training language models, it often struggles to capture global structure in reasoning tasks. Multi-token prediction (MTP) has recently emerged as a promising alternative, yet its underlying mechanisms remain poorly understood. In this paper, we study...
[]
null
43
2604.11912
title_snapshot
colm2026-0044
Resa: Efficient Reasoning Models via SAEs
null
[ "Shangshang Wang", "Julian Asilis", "Ömer Faruk Akgül", "Enes Burak Bilgin", "Ollie Liu", "Deqing Fu", "Willie Neiswanger" ]
null
null
null
[]
null
44
null
null
colm2026-0045
Simulating Organized Group Behavior: New Framework, Benchmark, and Analysis
https://arxiv.org/abs/2604.09874
[ "Xinkai Zou", "Yiming Huang", "Zhuohang Wu", "Jian Sha", "Nan Huang", "Longfei Yun", "Jingbo Shang", "Letian Peng" ]
null
null
Simulating how organized groups (e.g., corporations) make decisions (e.g., responding to a competitor's move) is essential for understanding real-world dynamics and could benefit relevant applications (e.g., market prediction). In this paper, we formalize this problem as a concrete research platform for group behavior ...
[]
null
45
2604.09874
title_snapshot
colm2026-0046
Uncertainty as a Planning Signal: Multi-Turn Decision Making for Goal-Oriented Conversation
https://arxiv.org/abs/2604.03924
[ "Xinyi Ling", "Ye Liu", "Reza Averly", "Xia Ning" ]
null
null
Goal-oriented conversational systems require making sequential decisions under uncertainty about the user's intent, where the algorithm must balance information acquisition and target commitment over multiple turns. Existing approaches address this challenge from different perspectives: structured methods enable multi-...
[]
null
46
2604.03924
title_snapshot
colm2026-0047
T-REX: Mixture-of-Rank-One-Experts with Semantic-aware Intuition for Multi-task Large Language Model Finetuning
https://arxiv.org/abs/2404.08985
[ "Rongyu Zhang", "Yijiang Liu", "Huanrui Yang", "Shenli Zheng", "Li Du", "Dan Wang", "Yuan Du", "Shanghang Zhang" ]
null
null
Large language models (LLMs) encounter significant adaptation challenges in diverse multitask finetuning. Mixture-of-experts (MoE) provides a promising solution with a dynamic architecture, enabling effective task decoupling. However, scaling up the number of MoE experts incurs substantial parameter and computational o...
[]
null
47
2404.08985
title_snapshot
colm2026-0048
Do LLMs Know What Is Private Internally? Probing and Steering Contextual Privacy Norms in Large Language Model Representations
https://arxiv.org/abs/2604.00209
[ "Haoran Wang", "Li Xiong", "Kai Shu" ]
null
null
Large language models (LLMs) are increasingly deployed in high-stakes settings, yet they frequently violate contextual privacy by disclosing private information in situations where humans would exercise discretion. This raises a fundamental question: do LLMs internally encode contextual privacy norms, and if so, why do...
[]
null
48
2604.00209
title_snapshot
colm2026-0049
Monocultural Biases: Correlated biases in large language models lead to unequal systemic exclusion rates in hiring
null
[ "Matthew Bramwell Bone", "Fabian Stephany", "R. Maria del Rio-Chanona" ]
null
null
null
[]
null
49
null
null
colm2026-0050
Safety Cost of Steering Vectors Is Separable and Reducible
null
[ "Yuxiao Li", "Gjergji Kasneci" ]
null
null
null
[]
null
50
null
null
colm2026-0051
Bagpiper: Solving Open-Ended Audio Tasks via Rich Captions
https://arxiv.org/abs/2602.05220
[ "Jinchuan Tian", "Haoran Wang", "Bo-Hao Su", "Chien-yu Huang", "Qingzheng Wang", "Jiatong Shi", "William Chen", "Xun Gong", "Siddhant Arora", "Chin-Jou Li", "Masao Someki", "Takashi Maekaku", "Yusuke Shinohara", "Jin Sakuma", "Keita Goto", "Chao-Han Huck Yang", "Shinji Watanabe" ]
null
null
Current audio foundation models typically rely on rigid, task-specific supervision, addressing isolated factors of audio rather than the whole. In contrast, human intelligence processes audio holistically, seamlessly bridging physical signals with abstract cognitive concepts to execute complex tasks. Grounded in this p...
[]
null
51
2602.05220
title_snapshot
colm2026-0052
Introspective Diffusion Language Models
https://arxiv.org/abs/2604.11035
[ "Yifan Yu", "Yuqing Jian", "Junxiong Wang", "Zhongzhu Zhou", "Donglin Zhuang", "Xinyu Fang", "Xiaoxia Wu", "Qingyang Wu", "Shuaiwen Leon Song", "Tri Dao", "Ben Athiwaratkun", "James Zou", "Fan Lai", "Chenfeng Xu" ]
null
null
Diffusion language models promise parallel generation, yet still lag behind autoregressive (AR) models in quality. We stem this gap to a failure of introspective consistency: AR models agree with their own generations, while DLMs often do not. We define the introspective acceptance rate, which measures whether a model ...
[]
null
52
2604.11035
title_snapshot
colm2026-0053
CoreSemDB: Benchmarking Hybrid Semantic-Relational Query Processing over Text-Rich Databases
null
[ "Yuchen Tian", "Jianxiong Guo", "Hao Zhang", "Jing Ma", "Congli Gao" ]
null
null
null
[]
null
53
null
null
colm2026-0054
A Self-Pruning Transformer: Extreme KV-Cache Compression with Universal Attention
null
[ "Davis Wertheimer", "Haochen Shen", "Ahan Gupta", "Derrick Liu", "Yu Chin Fabian Lim", "Mudhakar Srivatsa", "Raghu K. Ganti", "Minjia Zhang", "Naigang Wang" ]
null
null
null
[]
null
54
null
null
colm2026-0055
Credit Cards, Confusion, Computation, and Consequences: What Can We Uncover About Language Model Reasoning?
null
[ "Arnav Hiray", "Agam Shah", "Caleb Lu", "Meghaj Tarte", "Harsit Mittal", "Sudheer Chava" ]
null
null
null
[]
null
55
null
null
colm2026-0056
Learning Next Action Predictors from Human-Computer Interaction
https://arxiv.org/abs/2603.05923
[ "Omar Shaikh", "Valentin Teutschbein", "Kanishk Gandhi", "Yikun Chi", "Nick Haber", "Thomas N. Robinson", "Nilam Ram", "Byron Reeves", "Sherry Yang", "Michael S. Bernstein", "Diyi Yang" ]
null
null
Truly proactive AI systems must anticipate what we will do next. This foresight demands far richer information than the sparse signals we type into our prompts -- it demands reasoning over the entire context of what we see and do. We formalize this as next action prediction (NAP): given a sequence of a user's multimoda...
[]
null
56
2603.05923
title_snapshot
colm2026-0057
Human vs Machine Translation Detection: A Cross-Model, Cross-Domain, and Low-Resource Analysis
null
[ "Yabsera Yemanberhan", "Prasenjit Mitra", "Roald Eiselen" ]
null
null
null
[]
null
57
null
null
colm2026-0058
Generating Pretraining Tokens from Organic Data for Data-Bound Scaling
https://arxiv.org/abs/2605.17849
[ "Zichun Yu", "Chenyan Xiong" ]
null
null
LLM pretraining is shifting from a compute-bound to a data-bound regime, where available human (organic) text falls far short of scaling demands. However, reaching the data-bound regime does not mean the model has fully utilized its organic corpus. In this paper, we introduce SynPro, a synthetic data generation framewo...
[]
null
58
2605.17849
title_snapshot
colm2026-0059
Latent Structure of Affective Representations in Large Language Models
https://arxiv.org/abs/2604.07382
[ "Benjamin J. Choi", "Melanie Weber" ]
null
null
The geometric structure of latent representations in large language models (LLMs) is an active area of research, driven in part by its implications for model transparency and AI safety. Existing literature has focused mainly on general geometric and topological properties of the learnt representations, but due to a lac...
[]
null
59
2604.07382
title_snapshot
colm2026-0060
CritICL: Inference-Time Weak-to-Strong Generalization from Small Language Model Failure Modes
null
[ "Yufan Wu", "Zhengyi Hu", "Lang Wei", "Ruichen Li", "Qifan Yang", "Yinghui He", "Ting Zhu" ]
null
null
null
[]
null
60
null
null
colm2026-0061
Intern-S1-MO: Long-horizon Reasoning Agent for Olympiad-Level Mathematical Problem Solving
null
[ "Songyang Gao", "Yuzhe Gu", "Zijian Wu", "Lingkai Kong", "Wenwei Zhang", "ZhongruiCai", "Fan Zheng", "Tianyou Ma", "Junhao Shen", "Haiteng Zhao", "Duanyang Zhang", "Huilun Zhang", "Kuikun Liu", "Chengqi Lyu", "YanhuiDuan", "Chiyu Chen", "Ningsheng Ma", "Jianfei Gao", "Han Lyu", ...
null
null
null
[]
null
61
null
null
colm2026-0062
Alignment Whack-a-Mole : Finetuning Activates Verbatim Recall of Copyrighted Books in Large Language Models
https://arxiv.org/abs/2603.20957
[ "Xinyue Liu", "Niloofar Mireshghallah", "Jane C. Ginsburg", "Tuhin Chakrabarty" ]
null
null
Frontier LLM companies have repeatedly assured courts and regulators that their models do not store copies of training data. They further rely on safety alignment strategies via RLHF, system prompts, and output filters to block verbatim regurgitation of copyrighted works, and have cited the efficacy of these measures i...
[]
null
62
2603.20957
title_snapshot
colm2026-0063
Blind Refusal: Language Models Refuse to Help Users Evade Unjust, Absurd, and Illegitimate Rules
https://arxiv.org/abs/2604.06233
[ "Cameron Pattison", "Lorenzo Manuali", "Seth Lazar" ]
null
null
Safety-trained language models routinely refuse requests for help circumventing rules. But not all rules deserve compliance. When users ask for help evading rules imposed by an illegitimate authority, rules that are deeply unjust or absurd in their content or application, or rules that admit of justified exceptions, re...
[]
null
63
2604.06233
title_snapshot
colm2026-0064
Learning to Learn from Language Feedback with Social Meta-Learning
https://arxiv.org/abs/2602.16488
[ "Jonathan Cook", "Diego Antognini", "Martin Klissarov", "Claudiu Cristian Musat", "Edward Grefenstette" ]
null
null
Large language models (LLMs) often struggle to learn from corrective feedback within a conversational context. They are rarely proactive in soliciting this feedback, even when faced with ambiguity, which can make their dialogues feel static, one-sided, and lacking the adaptive qualities of human conversation. To addres...
[]
null
64
2602.16488
title_snapshot
colm2026-0065
Instruction-tuned LLMs for Full Argumentative Component Detection
null
[ "Sofiane Elguendouze", "Erwan Hain", "Elena Cabrio", "Serena Villata" ]
null
null
null
[]
null
65
null
null
colm2026-0066
YC-Bench: Benchmarking AI Agents for Long-Term Planning and Consistent Execution.
https://arxiv.org/abs/2604.01212
[ "Muyu He", "Adit Jain", "Anand Kumar", "Vincent Tu", "Soumyadeep Bakshi", "Sachin Patro", "Nazneen Rajani" ]
null
null
As LLM agents tackle increasingly complex tasks, a critical question is whether they can maintain strategic coherence over long horizons: planning under uncertainty, learning from delayed feedback, and adapting when early mistakes compound. We introduce $\texttt{YC-Bench}$, a benchmark that evaluates these capabilities...
[]
null
66
2604.01212
title_snapshot
colm2026-0067
TritonRL: Training LLMs to Think and Code Triton Without Cheating
https://arxiv.org/abs/2510.17891
[ "Jiin Woo", "Shaowei Zhu", "Allen Nie", "Zhen Jia", "Yida Wang", "Youngsuk Park" ]
null
null
The rapid evolution of Large Language Models (LLMs) has driven a growing demand for automated, high-performance system kernels to accelerate machine learning workloads. We introduce TritonRL, a domain-specialized 8B-scale LLM for Triton programming, trained via a novel reinforcement learning (RL) framework. While Trito...
[]
null
67
2510.17891
title_snapshot
colm2026-0068
Emergent Negligence: How Profit Mandates Induce Alignment Failures in LLMs
null
[ "Eric So" ]
null
null
null
[]
null
68
null
null
colm2026-0069
Language Models Might Not Understand You: Evaluating Theory of Mind via Story Prompting
https://arxiv.org/abs/2506.19089
[ "Nathaniel Getachew", "Abulhair Saparov" ]
null
null
We introduce StorySim, a programmable framework for synthetically generating stories to evaluate the theory of mind (ToM) and world modeling (WM) capabilities of large language models (LLMs). Unlike prior benchmarks that may suffer from contamination in pretraining data, or rely on an LLM for generation, StorySim produ...
[]
null
69
2506.19089
title_snapshot
colm2026-0070
Routing Entropy: A Hidden Self-Verifier for Free in Mixture-of-Experts LLMs
null
[ "Zhongyang Li", "Ziyue Li", "Tianyi Zhou" ]
null
null
null
[]
null
70
null
null
colm2026-0071
Discovering Hierarchical Latent Capabilities of Language Models from Observational Evaluation Data
null
[ "Jikai Jin", "Vasilis Syrgkanis", "Sham M. Kakade", "Hanlin Zhang" ]
null
null
null
[]
null
71
null
null
colm2026-0072
TierMem: Balancing Compressed Memory and Raw Evidence for Long-Horizon Agent Memory
null
[ "Qiming Zhu", "Shunian Chen", "Rui Yu", "Zhehao Wu", "Benyou Wang" ]
null
null
null
[]
null
72
null
null
colm2026-0073
Extended to Reality: Prompt Injection in 3D Environments
https://arxiv.org/abs/2602.07104
[ "Zhuoheng Li", "Ying Chen" ]
null
null
Multimodal large language models (MLLMs) have advanced the capabilities to interpret and act on visual input in 3D environments, empowering diverse applications such as robotics and situated conversational agents. When MLLMs reason over camera-captured views of the physical world, a new attack surface emerges: an attac...
[]
null
73
2602.07104
title_snapshot
colm2026-0074
WildTableBench: Benchmarking Multimodal Foundation Models on Table Understanding In the Wild
https://arxiv.org/abs/2605.01018
[ "Junzhe Huang", "Xiaoxiao Sun", "Yan Yang", "Yuxuan Hou", "Ruotian Zhang", "Sirui Li", "Hehe Fan", "Serena Yeung-Levy", "Xin Yu" ]
null
null
Using multimodal foundation models to analyze table images is a high-value yet challenging application in consumer and enterprise scenarios. Despite its importance, current evaluations rely largely on structured-text tables or clean rendered images, leaving the visual complexity of in-the-wild table images underexplore...
[]
null
74
2605.01018
title_snapshot
colm2026-0075
SciTaRC: Benchmarking QA on Scientific Tabular Data that Requires Language Reasoning and Complex Computation
https://arxiv.org/abs/2603.08910
[ "Hexuan Wang", "Yaxuan Ren", "Srikar Bommireddypalli", "Shuxian Chen", "Adarsh Prabhudesai", "Rongkun Zhou", "Elina Baral", "Philipp Koehn" ]
null
null
We introduce SciTaRC, an expert-authored benchmark of questions about tabular data in scientific papers requiring both deep language reasoning and complex computation. We show that current state-of-the-art AI models fail on at least 23% of these questions, a gap that remains significant even for highly capable open-wei...
[]
null
75
2603.08910
title_snapshot
colm2026-0076
Semantic Differentiation for Tackling Challenges in Watermarking Low-Entropy Constrained Generation Outputs
https://arxiv.org/abs/2601.11629
[ "Nghia T. Le", "Alan Ritter", "Kartik Goyal" ]
null
null
We demonstrate that while the current approaches for language model watermarking are effective for open-ended generation, they are inadequate at watermarking LM outputs for constrained generation tasks with low-entropy output spaces. Therefore, we devise SeqMark, a sequence-level watermarking algorithm with semantic di...
[]
null
76
2601.11629
title_snapshot
colm2026-0077
When Do LLMs Admit Their Mistakes? Understanding The Role Of Model Belief In Retraction
https://arxiv.org/abs/2505.16170
[ "Yuqing Yang", "Robin Jia" ]
null
null
Can large language models (LLMs) admit their mistakes when they should know better? In this work, we study when and why LLMs choose to retract, i.e., spontaneously and immediately acknowledge their errors. Using model-specific testbeds, we find that while LLMs are capable of retraction, they do so only rarely, even whe...
[]
null
77
2505.16170
title_snapshot
colm2026-0078
Barriers to Universal Reasoning with Transformers (and How to Overcome them)
https://arxiv.org/abs/2604.25800
[ "Oliver Kraus", "Yash Sarrof", "Yuekun Yao", "Alexander Koller", "Michael Hahn" ]
null
null
Chain-of-Thought (CoT) has been shown to empirically improve Transformers' performance, and theoretically increase their expressivity to Turing completeness. However, whether Transformers can learn to generalize to CoT traces longer than those seen during training is understudied. We use recent theoretical frameworks f...
[]
null
78
2604.25800
title_snapshot
colm2026-0079
Unable to Forget: Proactive Interference Reveals Working Memory Limits in LLMs Beyond Context Length
https://arxiv.org/abs/2506.08184
[ "Chupei Wang", "Jiaqiu Vince Sun" ]
null
null
Information retrieval in Large Language Models (LLMs) is increasingly recognized as intertwined with generation capabilities rather than mere lookup. While longer contexts are often assumed to improve retrieval, the effects of intra-context interference remain understudied. To address this, we adapt the proactive inter...
[]
null
79
2506.08184
title_snapshot
colm2026-0080
Restoring Generalization in Fine-tuned Multimodal LLMs via Geometric Alignment
null
[ "Lixuan He", "Shikang Zheng", "Guantao Chen" ]
null
null
null
[]
null
80
null
null
colm2026-0081
SWE-chat: Coding Agent Interactions From Real Users in the Wild
https://arxiv.org/abs/2604.20779
[ "Joachim Baumann", "Vishakh Padmakumar", "Xiang Li", "John Yang", "Diyi Yang", "Sanmi Koyejo" ]
null
null
AI coding agents are being adopted at scale, yet we lack empirical evidence on how people actually use them and how much of their output is useful in practice. We present SWE-chat, the first large-scale dataset of real coding agent sessions collected from open-source developers in the wild. The dataset currently contai...
[]
null
81
2604.20779
title_snapshot
colm2026-0082
VehicleMemBench: An Executable Benchmark for Multi-User Long-Term Memory in In-Vehicle Agents
https://arxiv.org/abs/2603.23840
[ "Yuhao Chen", "Yi Xu", "Xinyun Ding", "Xiang Fang", "Shuochen Liu", "Luxi Lin", "Qingyu Zhang", "Ya Li", "Quan Liu", "Tong Xu" ]
null
null
With the growing demand for intelligent in-vehicle experiences, vehicle-based agents are evolving from simple assistants to long-term companions. This evolution requires agents to continuously model multi-user preferences and make reliable decisions in the face of inter-user preference conflicts and changing habits ove...
[]
null
82
2603.23840
title_snapshot
colm2026-0083
ReFIne: A Framework for Trustworthy Large Reasoning Models with Reliability, Faithfulness, and Interpretability
https://arxiv.org/abs/2510.09062
[ "Chung-En Sun", "Ge Yan", "Akshay R. Kulkarni", "Tsui-Wei Weng" ]
null
null
Recent advances in long chain-of-thought (CoT) reasoning have largely prioritized answer accuracy and token efficiency, while overlooking aspects critical to trustworthiness. We argue that usable reasoning systems must be trustworthy, characterized by three properties: interpretability, faithfulness, and reliability. T...
[]
null
83
2510.09062
title_snapshot
colm2026-0084
The Devil Is in the Interface: Evaluating How Tool Architecture Shapes Coding Agent Behavior
null
[ "Xiangzhe Xu", "Hamidreza Saghir", "Qianhui Wu", "Marc-Alexandre Côté", "Tong Wang", "Kiran Lakkaraju", "Kexin Pei", "Xiangyu Zhang" ]
null
null
null
[]
null
84
null
null
colm2026-0085
Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges
https://arxiv.org/abs/2602.13576
[ "Ruomeng Ding", "Yifei Pang", "He Sun", "Yizhong Wang", "Steven Wu", "Zhun Deng" ]
null
null
Evaluation and alignment pipelines for large language models increasingly rely on LLM-based judges, whose behavior is guided by natural-language rubrics and validated on benchmarks. We identify a previously under-recognized vulnerability in this workflow, which we term Rubric-Induced Preference Drift (RIPD). Even when ...
[]
null
85
2602.13576
title_snapshot
colm2026-0086
DeepScholar-Bench: A Live Benchmark for Automated Evaluation of Generative Research Synthesis
null
[ "Liana Patel", "Negar Arabzadeh", "Harshit Gupta", "Ankita Sundar", "Alon Y. Halevy", "Ion Stoica", "Matei Zaharia", "Carlos Guestrin" ]
null
null
null
[]
null
86
null
null
colm2026-0087
ADAG: Automatically Describing Attribution Graphs
https://arxiv.org/abs/2604.07615
[ "Aryaman Arora", "Zhengxuan Wu", "Jacob Steinhardt", "Sarah Schwettmann" ]
null
null
In language model interpretability research, \textbf{circuit tracing} aims to identify which internal features causally contributed to a particular output and how they affected each other, with the goal of explaining the computations underlying some behaviour. However, all prior circuit tracing work has relied on ad-ho...
[]
null
87
2604.07615
title_snapshot
colm2026-0088
Lost in the Maze: Overcoming Context Limitations in Long-Horizon Agentic Search
https://arxiv.org/abs/2510.18939
[ "Howard Yen", "Yoonsang Lee", "Ashwin Paranjape", "Mengzhou Xia", "Thejas Venkatesh", "Jack Hessel", "Danqi Chen", "Yuhao Zhang" ]
null
null
Long-horizon agentic search requires iteratively exploring the web over long trajectories and synthesizing information across many sources, enabling powerful applications like deep research systems. In this work, we show that popular agentic search frameworks struggle to scale to long trajectories primarily due to cont...
[]
null
88
2510.18939
title_snapshot
colm2026-0089
VeriSoftBench: Repository-Scale Formal Verification Benchmarks for Lean
https://arxiv.org/abs/2602.18307
[ "Yutong Xin", "Jocelyn Qiaochu Chen", "Greg Durrett", "Isil Dillig" ]
null
null
Large language models have achieved striking results in interactive theorem proving, particularly in Lean. However, most benchmarks for LLM-based proof automation are drawn from mathematics in the Mathlib ecosystem, whereas proofs in software verification are developed inside definition-rich codebases with substantial ...
[]
null
89
2602.18307
title_snapshot
colm2026-0090
PolicyLong: Towards On-Policy Context Extension
https://arxiv.org/abs/2604.07809
[ "Junlong Jia", "Ziyang Chen", "Xing W", "Chaochen Gao", "TingHao YU", "feng zhang", "Songlin Hu", "Binghui Guo" ]
null
null
Extending LLM context windows is hindered by scarce high-quality long-context data. Recent methods synthesize data with genuine long-range dependencies via information-theoretic verification, selecting contexts that reduce a base model's predictive entropy. However, their single-pass offline construction with a fixed m...
[]
null
90
2604.07809
title_snapshot
colm2026-0091
Structure Before Collapse: Transient Semantic Geometry in Next-Token Prediction
https://arxiv.org/abs/2606.26749
[ "Yize Zhao", "Isabel Papadimitriou", "Christos Thrampoulidis" ]
null
null
Neural Collapse predicts that balanced one-hot classification pushes model representations to be equally far from each other; a symmetric configuration that depends only on the output label and ignores any semantic similarity in the inputs. This creates a puzzle: next-token prediction language models are trained predom...
[]
null
91
2606.26749
title_snapshot
colm2026-0092
MetaLint: Easy-to-Hard Generalization for Code Linting
https://arxiv.org/abs/2507.11687
[ "Atharva Naik", "Lawanya Baghel", "Dhatchinamoorthi Kunde Govindarajan", "Darsh Agrawal", "Yiqing Xie", "Daniel Fried", "Carolyn Rose" ]
null
null
Large language models excel at code generation but struggle with code linting, particularly in generalizing to unseen or evolving best practices beyond those observed during training. We introduce MetaLint, a meta-learning framework that formulates code linting as an instruction-following task, where a model evaluates ...
[]
null
92
2507.11687
title_snapshot
colm2026-0093
Path-Constrained Mixture-of-Experts
https://arxiv.org/abs/2603.18297
[ "Zijin Gu", "Tatiana Likhomanenko", "Vimal Thilak", "Jason Ramapuram", "Navdeep Jaitly" ]
null
null
Sparse Mixture-of-Experts (MoE) architectures route each token through a subset of experts at each layer independently. We propose viewing MoE computation through the lens of \emph{expert paths} -- the sequence of expert selections a token makes across all layers. This perspective reveals that, despite $N^L$ possible p...
[]
null
93
2603.18297
title_snapshot
colm2026-0094
Beyond Semantics: Rediscovering Spatial Awareness in Vision-Language Models
https://arxiv.org/abs/2503.17349
[ "Jianing Qi", "Jiawei Liu", "Hao Tang", "Zhigang Zhu" ]
null
null
Vision Language Models (VLMs) excel at identifying and describing objects but often fail at spatial reasoning. We study why VLMs, such as LLaVA, underutilize spatial cues despite having positional encodings and spatially rich vision encoder features. Our analysis reveals a key imbalance: vision token embeddings have mu...
[]
null
94
2503.17349
title_snapshot
colm2026-0095
What AI Benchmarks Actually Measure: Adapting Convergent and Discriminant Validity to Interrogate Fifty-Six AI Benchmarks
null
[ "Meera Desai", "Sang T. Truong", "Hanna Wallach", "Alex Chouldechova", "A. Feder Cooper", "Jean Garcia-Gathright", "Daniel E. Ho", "Abigail Z. Jacobs", "Sanmi Koyejo", "Nicholas Pangakis", "Angelina Wang" ]
null
null
null
[]
null
95
null
null
colm2026-0096
Understanding Machine Unlearning Through the Lens of Mode Connectivity
https://arxiv.org/abs/2504.06407
[ "Jiali Cheng", "Hadi Amiri" ]
null
null
Machine Unlearning aims to remove undesired information from trained models without requiring full retraining from scratch. Despite recent advancements, their underlying loss landscapes and optimization dynamics received less attention. In this paper, we investigate and analyze machine unlearning through the lens of ...
[]
null
96
2504.06407
title_snapshot
colm2026-0097
Off-Policy Token-Level Labels Improve Out-of-Distribution Generalization for Summarization
null
[ "Zitong Huang", "Gustavo Adolpho Lucas de Carvalho", "Deqing Fu", "Robin Jia" ]
null
null
null
[]
null
97
null
null
colm2026-0098
Neuro-symbolic PRM: Enhancing Scientific Reasoning via Structured Traces and Symbolic Verification
null
[ "Yuxin Zi", "Cong Xu", "Suparna Bhattacharya", "Martin Foltin", "Amit Sheth" ]
null
null
null
[]
null
98
null
null
colm2026-0099
TRIDENT: Benchmarking LLM Safety in Finance, Medicine, and Law
https://arxiv.org/abs/2507.21134
[ "Zheng Hui", "Yijiang River Dong", "Ehsan Shareghi", "Nigel Collier" ]
null
null
As large language models (LLMs) are increasingly deployed in high-risk domains such as law, finance, and medicine, systematically evaluating their domain-specific safety and compliance becomes critical. While prior work has largely focused on improving LLM performance in these domains, it has often neglected the evalua...
[]
null
99
2507.21134
title_snapshot
colm2026-0100
Studying the Soupability of Documents in State Space Models
https://arxiv.org/abs/2505.24033
[ "Yasaman Jafari", "Zixian Wang", "Leon Bergen", "Taylor Berg-Kirkpatrick" ]
null
null
We investigate whether hidden states from Structured State Space Models (SSMs) can be merged post hoc to support downstream reasoning. Inspired by model souping, we study document souping, a strategy where documents are encoded independently, and their representations are pooled, via simple operations like averaging, i...
[]
null
100
2505.24033
title_snapshot
End of preview. Expand in Data Studio
README.md exists but content is empty.
Downloads last month
32