AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents

TL;DR

AIRS-Bench is a comprehensive suite of 20 diverse tasks designed to evaluate and advance AI agents' capabilities across the entire scientific research process, highlighting current strengths and areas for improvement.

Contribution

Introduces AIRS-Bench, a versatile, open-source benchmark for assessing AI agents in scientific research tasks across multiple domains and research stages.

Findings

Agents outperform humans in 4 tasks

Agents do not reach theoretical performance limits

Significant room for improvement remains

Abstract

LLM agents hold significant promise for advancing scientific research. To accelerate this progress, we introduce AIRS-Bench (the AI Research Science Benchmark), a suite of 20 tasks sourced from state-of-the-art machine learning papers. These tasks span diverse domains, including language modeling, mathematics, bioinformatics, and time series forecasting. AIRS-Bench tasks assess agentic capabilities over the full research lifecycle -- including idea generation, experiment analysis and iterative refinement -- without providing baseline code. The AIRS-Bench task format is versatile, enabling easy integration of new tasks and rigorous comparison across different agentic frameworks. We establish baselines using frontier models paired with both sequential and parallel scaffolds. Our results show that agents exceed human SOTA in four tasks but fail to match it in sixteen others. Even when agents surpass human benchmarks, they do not reach the theoretical performance ceiling for the underlying tasks. These findings indicate that AIRS-Bench is far from saturated and offers substantial room for improvement. We open-source the AIRS-Bench task definitions and evaluation code to catalyze further development in autonomous scientific research.