Can Coding Agents Reproduce Findings in Computational Materials Science?

2026-05-01

AI Summary

This work addresses the reproducibility challenges in computational materials science, which stem from incomplete workflows, complex toolchains, and difficulties in validation. To systematically evaluate the capability of large language model–driven coding agents in reconstructing and executing computational pipelines to verify scientific claims, the authors introduce AutoMat—the first end-to-end benchmark tailored to this domain. AutoMat integrates expert-curated workflows, domain-specific toolchains, and an automated execution environment, assessing agent performance across three dimensions: workflow reconstruction, tool invocation, and scientific validation. Experimental results reveal that even state-of-the-art agents achieve only a 54.1% success rate, primarily hindered by missing procedural details, methodological biases, and execution fragility, thereby exposing critical bottlenecks in autonomous AI systems for scientific discovery.

Abstract

Large language models are increasingly deployed as autonomous coding agents and have achieved remarkably strong performance on software engineering benchmarks. However, it is unclear whether such success transfers to computational scientific workflows, where tasks require not only strong coding ability, but also the ability to navigate complex, domain-specific procedures and to interpret results in the context of scientific claims. To address this question, we present AutoMat, a benchmark for evaluating LLM-based agents' ability to reproduce claims from computational materials science. AutoMat poses three interrelated challenges: recovering underspecified computational procedures, navigating specialized toolchains, and determining whether the resulting evidence supports a claim. By working closely with subject matter experts, we curate a set of claims from real materials science papers to test whether coding agents can recover and execute the end-to-end workflow needed to support (or undermine) such claims. We then evaluate multiple representative coding agent settings across several foundation models. Our results show that current LLM-based agents obtain low overall success rates on AutoMat, with the best-performing setting achieving a success rate of only 54.1%. Error analysis further reveals that agents perform worst when workflows must be reconstructed from paper text alone and that they fail primarily due to incomplete procedures, methodological deviations, and execution fragility. Taken together, these findings position AutoMat as both a benchmark for computational scientific reproducibility and a tool for diagnosing the current limitations of agentic systems in AI-for-science settings.

Problem

• computational materials science
• reproducibility
• coding agents
• scientific workflows
• large language models

Innovation

• coding agents
• computational reproducibility
• materials science
• LLM benchmark
• scientific workflows