Hugging Face's logo

Datasets:
FDAbench2026
/
FDABench-File

ArXiv:
License:
Dataset card Files
xet
 Community

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.

By requesting access, you agree to the following terms:

  1. Academic, non-commercial research use only. This corpus must not be used to
    train, fine-tune, or evaluate commercial products or services without separate
    permission from the underlying rights holders.

  2. Per-source licensing applies. Files in this dataset retain the licenses
    of their original sources (arXiv non-exclusive, Public Domain, CC-BY-NC-SA,
    and "academic-use-only" for items whose licenses could not be conclusively
    determined). See manifest.csv for per-file license tags and LICENSE.md
    for the full text.

  3. Attribution required. Any redistribution of derived work must credit the
    original sources alongside the FDABench paper.

  4. No redistribution of the raw bundle. You may release derivative artifacts
    (embeddings, chunks, evaluation snapshots) but not re-host the raw files
    bundled here.

  5. Takedown policy. If you are a rights holder and want a file removed,
    email FDAbench2026@gmail.com with the path; we will remove within 7 days.

Log in or Sign Up to review the conditions and access this dataset content.

FDABench-File

Raw multimodal corpus distributed alongside the FDABench benchmark (paper · code · FDAbench-Full).

The corpus contains 1,418 PDFs, 246 images, 108 audio clips, 81 video clips, plus Whisper transcripts and a manifest, organized into 50 domain categories. Curated from publicly accessible sources: arXiv preprints, government and institutional publications, and open-licensed educational content. Detailed per-file licensing is in manifest.csv and LICENSE.md.

Terms of Use

This dataset is released for academic, non-commercial research only. Each file retains the license of its original source; the license column of manifest.csv records one of:

Tag Meaning Approx. count
arxiv-non-exclusive arXiv preprint, redistributable under arXiv's non-exclusive license 597
public-domain Public domain (LibriVox audiobooks, government publications) 12
cc-by-nc-sa Creative Commons BY-NC-SA (e.g. MIT OpenCourseWare) 9
academic-use-only License could not be conclusively determined from source metadata; treat as fair-use academic-only and contact original source for any non-academic use 1,285

Items in the academic-use-only bucket are bundled in good faith from publicly accessible sources for the sole purpose of academic benchmark reproducibility. We do not assert any rights to these files, and we do not authorize commercial redistribution.

Takedown

If you are the rights holder of any item and want it removed, email FDAbench2026@gmail.com with the file path (e.g. Energy_Utilities/foo.mp3). We will remove the file from main within 7 days of receipt.

Loading 

# Whole tree
from huggingface_hub import snapshot_download
snapshot_download("FDAbench2026/FDABench-File", repo_type="dataset",
                  local_dir="./Vector_Database")

# Single file
from huggingface_hub import hf_hub_download
path = hf_hub_download("FDAbench2026/FDABench-File",
                       "Genomics_Bioinformatics/000003.jpg",
                       repo_type="dataset")

Citation 

@inproceedings{wang2026fdabench,
  title     = {FDABench: A Benchmark for Data Agents on Analytical Queries over Heterogeneous Data},
  author    = {Wang, Ziting and Zhang, Shize and Yuan, Haitao and Zhu, Jinwei and Dong, Wei and Cong, Gao},
  booktitle = {Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD)},
  year      = {2026},
  note      = {To appear}
}
Downloads last month
197
Total file size:
11.5 GB

Paper for FDAbench2026/FDABench-File