Can Coding Agents Reproduce Findings in Computational Materials Science?

• URL: http://arxiv.org/abs/2605.00803v1
• Date: Fri, 01 May 2026 17:42:12 GMT
• Title: Can Coding Agents Reproduce Findings in Computational Materials Science?
• Abstract summary: We present AutoMat, a benchmark for evaluating large language models' ability to reproduce scientific claims.<n>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.<n>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%.
• Score: 49.254975563645786
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• 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.

Related papers

• From Atomic to Agentic: Towards Interpretable Evaluation of LLMs' Agentic Mathematical Capabilities [103.60384544335072]
  Large Language Models (LLMs) are evolving from performing end-to-end mathematical reasoning to integrating agentic intelligence.<n>Our framework aligns problem-solving agentic behaviors with a structured taxonomy of reusable mathematical atomic capabilities.<n> Experiments reveal that models with similar end-to-end accuracy can exhibit markedly different agentic capability profiles.
  arXiv  Detail & Related papers  (2026-08-27T10:52:06Z)
• Benchmarking AI Agents for Addressing Scientific Challenges Across Scales [118.2204632627895]
  SciAgentArena is a systematic benchmark for evaluating AI agents in real-world scientific research scenarios.<n>It comprises approximately 200 tasks with stepwise verification and an interactive, agent-agnostic environment for assessing diverse AI agents.<n>We find that current agents can contribute effectively to well-specified data-analysis, particularly when the task structure and evaluation criteria are clear.
  arXiv  Detail & Related papers  (2026-06-10T22:55:30Z)
• AI Coding Agents Can Reproduce Social Science Findings [5.410553210073652]
  Recent evidence suggests that AI coding agents can reproduce published findings when provided with original data and code.<n>We introduce SocSci-Repro-Bench, a benchmark of 221 tasks spanning four disciplines and 13 substantive domains.<n>We find that both can reproduce a large share of social science findings, with Claude Code substantially outperforming Codex.
  arXiv  Detail & Related papers  (2026-06-09T21:00:05Z)
• SCICONVBENCH: Benchmarking LLMs on Multi-Turn Clarification for Task Formulation in Computational Science [3.9311288356229057]
  Large Language Models (LLMs) are increasingly deployed as scientific AI as- sistants.<n>We introduce SCICONVBENCH, a benchmark for multi- turn clarification in scientific task formulation.<n>Our benchmark pairs a structured task with a rubric-based evaluation framework.
  arXiv  Detail & Related papers  (2026-05-18T16:34:45Z)
• Collider-Bench: Benchmarking AI Agents with Particle Physics Analysis Reproduction [0.02446672595462589]
  We introduce Collider-Bench, a benchmark for evaluating whether language-model agents can reproduce experimental analyses from the Large Hadron Collider (LHC) using only public papers and open scientific software.<n>Agents must therefore rely on physical reasoning, domain knowledge, and trial-and-error to fill these gaps.<n>Our results show that on average no agent reliably beats the physicist-in-the-loop solution.
  arXiv  Detail & Related papers  (2026-05-13T18:00:00Z)
• Agent psychometrics: Task-level performance prediction in agentic coding benchmarks [24.348135523715815]
  We present a framework for predicting success or failure on individual tasks tailored to the agentic coding regime.<n>Our approach augments Item Response Theory (IRT) with rich features extracted from tasks, including issue statements, repository contexts, solutions, and test cases.
  arXiv  Detail & Related papers  (2026-04-01T07:59:59Z)
• AI-for-Science Low-code Platform with Bayesian Adversarial Multi-Agent Framework [4.782965804438204]
  Large Language Models (LLMs) demonstrate potentials for automating scientific code generation but face challenges in reliability, error propagation, and evaluation.<n>We present a Bayesian adversarial multi-agent framework specifically designed for AI for Science (AI4S) tasks in the form of a Low-code Platform (LCP)<n>Three LLM-based agents are coordinated under the Bayesian framework: a Task Manager that structures user inputs into actionable plans and adaptive test cases, a Code Generator that produces candidate solutions, and an Evaluator providing comprehensive feedback.
  arXiv  Detail & Related papers  (2026-03-03T18:25:00Z)
• ReplicatorBench: Benchmarking LLM Agents for Replicability in Social and Behavioral Sciences [19.81372090301296]
  ReplicatorBench is an end-to-end benchmark for evaluating AI agents in research replication across three stages.<n>We develop ReplicatorAgent, an agentic framework equipped with necessary tools, like web search and iterative interaction with sandboxed environments.<n>We evaluate ReplicatorAgent across four underlying large language models (LLMs), as well as different design choices of programming language and levels of code access.
  arXiv  Detail & Related papers  (2026-02-11T20:42:10Z)
• Automating Computational Reproducibility in Social Science: Comparing Prompt-Based and Agent-Based Approaches [0.3637090866781549]
  Reproducing computational research is often assumed to be as simple as rerunning the original code with provided data.<n>This study investigates whether large language models and AI agents can automate the diagnosis and repair of such failures.
  arXiv  Detail & Related papers  (2026-02-09T11:59:59Z)
• An Agentic Framework for Autonomous Materials Computation [70.24472585135929]
  Large Language Models (LLMs) have emerged as powerful tools for accelerating scientific discovery.<n>Recent advances integrate LLMs into agentic frameworks, enabling retrieval, reasoning, and tool use for complex scientific experiments.<n>Here, we present a domain-specialized agent designed for reliable automation of first-principles materials computations.
  arXiv  Detail & Related papers  (2025-12-22T15:03:57Z)
• Simple Agents Outperform Experts in Biomedical Imaging Workflow Optimization [69.36509281190662]
  Adapting production-level computer vision tools to bespoke scientific datasets is a critical "last mile" bottleneck.<n>We consider using AI agents to automate this manual coding, and focus on the open question of optimal agent design.<n>We demonstrate that a simple agent framework consistently generates adaptation code that outperforms human-expert solutions.
  arXiv  Detail & Related papers  (2025-12-02T18:42:26Z)
• Toward Automated and Trustworthy Scientific Analysis and Visualization with LLM-Generated Code [6.068120728706316]
  Large language models (LLMs) offer a promising solution by generating code from natural language descriptions.<n>We construct a benchmark suite of domain-inspired prompts that reflect real-world research tasks.<n>Our findings show that, without human intervention, the reliability of LLM-generated code is limited.
  arXiv  Detail & Related papers  (2025-11-26T21:27:03Z)
• AutoMind: Adaptive Knowledgeable Agent for Automated Data Science [70.33796196103499]
  Large Language Model (LLM) agents have shown great potential in addressing real-world data science problems.<n>Existing frameworks depend on rigid, pre-defined and inflexible coding strategies.<n>We introduce AutoMind, an adaptive, knowledgeable LLM-agent framework.
  arXiv  Detail & Related papers  (2025-06-12T17:59:32Z)
• DrugPilot: LLM-based Parameterized Reasoning Agent for Drug Discovery [54.79763887844838]
  Large language models (LLMs) integrated with autonomous agents hold significant potential for advancing scientific discovery through automated reasoning and task execution.<n>We introduce DrugPilot, a LLM-based agent system with a parameterized reasoning architecture designed for end-to-end scientific in drug discovery.<n>DrugPilot significantly outperforms state-of-the-art agents such as ReAct and LoT, achieving task completion rates of 98.0%, 93.5%, and 64.0% for simple, multi-tool, and multi-turn scenarios, respectively.
  arXiv  Detail & Related papers  (2025-05-20T05:18:15Z)
• MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation [96.71370747681078]
  We introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate.
  arXiv  Detail & Related papers  (2023-10-05T04:06:12Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.