EvoMaster: A Foundational Evolving Agent Framework for Agentic Science at Scale

arXiv 2026 73.6 method

TLDR

EvoMaster is a domain-agnostic, self-evolving agent framework for scalable agentic science, achieving state-of-the-art results on four benchmarks.

Reasoning

The paper introduces a novel evolving agent framework that iteratively refines hypotheses and accumulates knowledge, with strong empirical results across multiple benchmarks. However, the abstract lacks discussion of limitations and only compares against a single baseline, limiting the depth of validation.

Read-first score

Read-first score 73.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 72.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=code,github

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark,evaluation,experiment

Topical relevance 42%

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Field roles

FrontierMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 25.

Keyword Scores

automated scientific discovery

scientific discovery agent

autonomous research agent

automated research

AI for scientific research

research automation

AI scientist

automated experimentation

experiment design agent

literature review agent

survey generation

paper writing agent

Deep Analysis

Innovations

Continuous self-evolution mechanism enabling agents to iteratively refine hypotheses, self-critique, and accumulate knowledge across experimental cycles
Domain-agnostic foundational framework that allows building self-evolving scientific agents for arbitrary disciplines in approximately 100 lines of code
SciMaster ecosystem instantiated across machine learning, physics, and general science domains

Methodology

EvoMaster is a foundational evolving agent framework that empowers agents to continuously self-evolve by refining hypotheses, self-critiquing, and accumulating knowledge. It is domain-agnostic and was used to build the SciMaster ecosystem, evaluated on four benchmarks (Humanity's Last Exam, MLE-Bench Lite, BrowseComp, FrontierScience) against the OpenClaw baseline.

Key Results

EvoMaster achieves state-of-the-art scores of 41.1%, 75.8%, 73.3%, and 53.3% on Humanity's Last Exam, MLE-Bench Lite, BrowseComp, and FrontierScience respectively, with relative improvements over OpenClaw ranging from +159% to +316%.

Resources

Paper Code