ScienceClaw: Benchmarking Continual Self-Evolution of AI-for-Science Agents Across the Natural and Social Sciences

Mingda Zhang, Wenjin Liu, Tiesunlong Shen, Zikai Xiao, Zhenghong Lin +4 more
10/6/2026
cs.AI

Official code: beita6969/ScienceClaw (906 stars, MIT, last commit 2026-10-06).

Abstract

Large language model agents are accelerating scientific automation, yet verified executions rarely become persistent program-level improvements, and existing evaluations do not examine this process across sequential tasks in both the natural and social sciences. We formalize ScienceClaw as fixed-parameter program self-evolution that unifies task solving, scientific verification, and program updates. ScienceClaw-Eval spans 23 disciplines and measures scientific correctness, evolutionary gain, retention, cross-dataset transfer, and evolution cost through sequential streams and independent reset evaluation. Our framework repairs executable workflows through multi-turn interaction, converts re-execution-verified failure--success trajectories into linked Skill and Operator candidates, and retains an update only when source-task replay reproduces the repair and independent scientific tasks improve. Code is available at https://github.com/beita6969/ScienceClaw.

View on arXivView PDF

Code Implementations(1)

beita6969/ScienceClawOfficial100%
MIT

Verifiable program-level self-evolution for AI-for-Science agents: Skills and typed Operators grow from replay-verified executions.

906105Go, CSSMar 11, 20262 weeks agoMIT
aiai-agentbioinformaticsliterature-reviewllm+9 more

Similar Papers

L3M: A Verifiable Harness and Runtime for Self-Extending and Self-Improving Agents

86% match2026

From Agent Loops to Deterministic Graphs: Execution Lineage for Reproducible AI-Native Work

85% match2026

Mimosa Framework: Toward Evolving Multi-Agent Systems for Scientific Research

84% match2026

TraceML: An Empirical Analysis of Human-Agent Planning in Machine Learning Development

84% match2026

PairCoder++: Pair Programming as a Universal Paradigm for Verified Code-Driven Multimodal and Structured-Artifact Generation

83% match2026

CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery

83% match1 repo7712026

Cite this paper

@article{zhang2026scienceclaw,
  title  = {ScienceClaw: Benchmarking Continual Self-Evolution of AI-for-Science Agents Across the Natural and Social Sciences},
  author = {Mingda Zhang and Wenjin Liu and Tiesunlong Shen and Zikai Xiao and Zhenghong Lin and Qing Xu and Erik Cambria and Xiaoying Tang and Haoran Luo},
  year   = {2026},
  eprint = {2610.08691},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2610.08691v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2610.08691.svg)](https://codeofpaper.com/paper/2610.08691)

Discussion