AI Paper Atlas

Research map / Large language models / Agents, code and math

ABC-Bench: An Agentic Bio-Capabilities Benchmark for Biosecurity

Andrew Bo Liu, Samira Nedungadi, Bryce Cai, Alex Kleinman, Harmon Bhasin, Seth Donoughe

ICML 2026 · Large language models › Agents, code and math · approach: Agent instruction-following benchmarks · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
compare frontier models to human expert baseliners on Fragment Designfrontier LLM agents · ABC-Bench Fragment Designmean score: 0.33 ± 0.12 -> 1.00 ± 0.00Improved
compare frontier models to human baseliners on Liquid Handling Robotfrontier LLM agents · ABC-Bench Liquid Handling Robotmean score: 0.20 ± 0.09 -> 1.00 ± 0.00Improved
compare models to human baseliners on Screening EvasionGemini 3.1 Pro Preview · ABC-Bench Screening Evasionmean score: 0.22 ± 0.07 -> 0.78 ± 0.02Improved
newer Claude Opus 4.6 versus Claude Opus 4Claude Opus · ABC-Bench Fragment Designmean score: 0.71 ± 0.05 -> 1.00 ± 0.00Improved
switch from creative Screening Evasion to well-documented Fragment DesignQwen3.5 397B-A17B · ABC-Benchmean score: 0.57 ± 0.02 -> 0.76 ± 0.02Improved
Screening Evasion prompt concealing malicious intentClaude Sonnet 4.6, Opus 4.6, GPT-5.4 · ABC-Bench Screening Evasionrefusal rate: 1.00Hurt
model refuses Fragment Design samplesClaude Opus 4 · ABC-Bench Fragment Designrefusal rate: 0.52Mixed
LLM-guided real wet-lab Gibson assembly on OpenTrons FlexGPT-o4-mini-high · Wet-lab DNA assemblysuccessful assemblies: 3 of 3Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy