scBench

An agent benchmark of verifiable scRNA-seq analysis problems; inspect_evals ships a 30-task public canonical slice, while the paper describes 394 held-out tasks.

Also known as: scBench: Evaluating AI Agents on Single-Cell RNA-seq Analysis, scbench

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categoryagentic
Subcategoryagentic single-cell RNA-seq analysis on .h5ad snapshots with deterministic graders
Page statusactive
Metricaccuracy (fraction of tasks passed); stderr also reported
Directionhigher_is_better
Unit%
Dataset size394
Dataset licenceApache-2.0
PublisherLatchBio

What it measures

scBench (LatchBio) asks an agent to analyse a real single-cell RNA-seq snapshot stored as AnnData `.h5ad` and write structured answers to `eval_answer.json`. Tasks cover quality control, normalization, dimensionality reduction, clustering, cell typing, differential expression, and trajectory analysis across sequencing platforms. Graders are deterministic (numeric tolerance, multiple choice, marker-gene precision/recall, label-set Jaccard, distribution comparison). The agent must load the data; a memorised textbook answer fails if it never touches the file. inspect_evals, the census harness for this id, runs the public canonical subset (30 tasks, five platforms), not the full paper set.

Task format

Agentic code execution in a sandbox. inspect_evals defaults to mini-SWE-agent via inspect-swe inside Docker, with a react()+bash fallback. Timeout defaults: 600s task, 300s bash. Answers are read only from eval_answer.json; there is no text-parsing fallback.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub