OmicsBench tests AI agents on long complex bioinformatics tasks
OmicsBench: a long-horizon multi-turn benchmark of AI agents on research-study-scale bioinformatics
Artificial Intelligence
Summary
AI tools are often asked to help with complicated biology studies, but most tests only check if they can do a single simple job. The authors created OmicsBench, a test with many linked steps, to better see how AI handles longer, more connected tasks. They found AI tools often struggle more when tasks depend on earlier answers, but giving feedback helps. They also made a way to find common AI mistakes, showing how AIs sometimes ignore instructions or guess answers from memory.
What this means in practice
- •For bioinformatics platform developers: Improve AI assistants to better handle multi-step bioinformatics workflows by using OmicsBench for testing and feedback integration.
- •For ai tool integrators in healthcare: Use OmicsBench to identify and reduce frequent AI errors when applying bioinformatics analyses that rely on dependent reasoning.
Authors
Gideon Shaked, Siddharth Nagisetty
Abstract
Computational biologists increasingly delegate substantial analyses to AI agents in long-horizon, multi-turn sessions, yet most benchmarks of agents' bioinformatics capabilities pose a single short task with a single-shot prompt. We present OmicsBench, a long-horizon, multi-turn bioinformatics benchmark for AI agents. Each of its 97 tasks links dependent subtasks into a directed acyclic graph, for 662 underlying subtasks in all. Each subtask is posed as one user turn and graded deterministically against an answer key, with optional in-context feedback on incorrect answers. Across 12 models on three harnesses, the slowest agent averages 3.6 hours per task, and no combination comes near saturating the benchmark, with mean scores of 0.43 to 0.57 out of a possible 1 without in-context feedback and 0.54 to 0.67 with it. Multi-turn evaluation also uncovers behavior that single-shot prompts cannot, as agents score lower on subtasks that depend on earlier subtasks, and in-context feedback helps agents recover some of that loss. Because deterministic scoring gives no qualitative account of how an agent failed, we also introduce an unsupervised failure mode discovery pipeline that describes, embeds, and clusters agents' failed attempts to surface emergent failure modes. Applied to OmicsBench, this pipeline exposes qualitative shortcomings in how agents carry out bioinformatics analyses, ranging from agents applying their own statistical filters in place of the specified ones to agents writing whole answers from memory.