#9 · ML / AI methods
LLM agents are being benchmarked on multistage biological data analysis tasks
The dominant thread across these items is evaluation and automation of biological analysis by LLM-based agents. Prompt-to-Paper grounds manuscript claims in retrieved literature and executes bioinformatics experiments with quality scoring. GeneBench-Pro and BAISBench test agents on statistical reasoning and on real single-cell transcriptomic discovery, with GeneBench-Pro reporting a best pass rate of 31.5% across 129 problems. Agentomics autonomously develops biomedical ML solutions across 20 datasets, including protein engineering. What is new is that agent execution itself is now the object of measurement, with reported results indicating substantial remaining gaps. These items share an agentic focus but differ in whether they build systems or evaluate them.
Published papers
Benchmarking AI scientists for omics data-driven biological discovery.
BAISBench benchmarks AI scientists on real single-cell transcriptomic data via cell type annotation across 15 datasets and 193 discovery questions from 41…
Agentomics: an agentic system that autonomously develops novel state-of-the-art solutions for biomedical machine learning tasks.
Agentomics is an LLM-powered agentic system that automates end-to-end biomedical ML experimentation, benchmarked on 20 datasets across protein engineering,…
Preprints
Prompt-to-Paper: Agentic AI System for Bioinformatics
Prompt-to-Paper is a multi-agent LLM framework that grounds manuscript claims in retrieved literature, executes bioinformatics experiments, and scores quality…
Topic history
| Date | Rank |
|---|---|
| #9 | |
| #24 | |
| #25 | |
| #34 |