Toward Engineering AGI: Benchmarking the Engineering Design Capabilities of LLMs
Paper • 2509.16204 • Published
You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset is released for research use. Access is reviewed and granted manually by the maintainers. Please state your name, affiliation, and intended use.
Log in or Sign Up to review the conditions and access this dataset content.
ED-D08
Engineering-design tasks reformatted into the unified smart-manufacturing SFT schema. Category E (code / program), task T-E2.
The repository name is an internal code. See Provenance below for the underlying dataset.
Records
101 records (no train/val/test split).
Images & attachments. The single representative input figure is inlined via the HF Image feature (original path kept in metadata.image_path); a images column (list[Image]) byte-inlines every genuine input figure, so multi-figure tasks (e.g. 6 noise renders, 8 membership-function plots) that one image cannot hold still travel in full (16 tasks, 25 figures). Answer/render figures are not shipped — they are listed as provenance paths in metadata.solution_figures / other_figures. Text attachments (.json, .vcd) are inlined into metadata.attachments_text (relative path → file content); genuine binary attachments (.npy arrays, .mp4 animations) ship as files under attachments/ and are listed in metadata.attached_files.
Unified SFT schema (8 fields)
| field | type | meaning |
|---|---|---|
query |
str | the question / query / instruction |
image |
Image | null | the single representative INPUT figure, byte-inlined (14 records); null otherwise. Original relative path kept in metadata.image_path |
images |
list[Image] | D08-specific field. Byte-inlined list of all genuine INPUT figures for the task (covers multi-figure tasks the single image cannot hold); empty list when the task has no input figure |
annot |
str | list[str] | label / answer / annotation (resolves solution.txt → solution.py → output-structure spec; never the prompt) |
reasoning |
str | null | always null here (solution text is not a CoT trace) |
cate |
"A".."E" | one of the five SFT categories (this dataset: E) |
task |
"T-xx" | unified task id (this dataset: T-E2) |
metadata |
str (JSON) | all other info: image_path, input_figures/solution_figures/other_figures, attached_files, attachments_text, annot_source, … |
Load
from datasets import load_dataset
ds = load_dataset("AI4Manufacturing/ED-D08")
Gated — request access on the dataset page; access is granted manually by the maintainers.
Provenance & license
This dataset is a reformatted derivative (unified SFT schema) of:
EngDesign — Toward Engineering AGI (NeurIPS 2025 Datasets & Benchmarks).
- Paper: https://arxiv.org/abs/2509.16204
- Code: https://github.com/agi4engineering/EngDesign
- Original data: https://huggingface.co/datasets/opt1zer/EngDesign
Refer to the upstream source for the original licensing terms; this reformatted version is shared for research use. Please cite the upstream work.
- Downloads last month
- 39