retroam/scbench-data

scBench Canonical Data Files

This dataset contains the .h5ad data files for the scBench canonical subset (30 evaluations across 5 platforms).

About scBench

scBench is a benchmark for agentic single-cell RNA-seq analysis. It evaluates whether AI agents can solve practical bioinformatics tasks with deterministic grading.

• Paper: https://arxiv.org/abs/2602.09063
• GitHub: https://github.com/latchbio/scbench
• Inspect Evals: https://github.com/UKGovernmentBEIS/inspect_evals

Files

This dataset includes 18 .h5ad files corresponding to the canonical subset:

• Platforms: Chromium, CSGenetics, Illumina, MissionBio, ParseBio
• Task categories: QC, normalization, dimensionality reduction, clustering, cell typing, differential expression, trajectory analysis

Usage

These files are used with the scBench evaluation in inspect_evals:

pip install inspect-evals
uv run inspect eval inspect_evals/scbench --model anthropic/claude-haiku-4-5-20251001

License

Apache 2.0 (following the scBench upstream license)

Citation

If you use this dataset, please cite the scBench paper:

@misc{scbench2025,
  title={scBench: A Benchmark for Agentic Single Cell Data Analysis},
  author={LatchBio},
  year={2025},
  eprint={2602.09063},
  archivePrefix={arXiv}
}