Toward Auditable AI Scientists: A Hypothesis Evolution Protocol for LLM Agents

arXiv 2026 54.1 method

TLDR

Proposes Hypothesis Evolution Protocol (HEP) for LLM agents to make hypothesis generation, evaluation, and evolution auditable in scientific discovery.

Reasoning

The paper addresses a clear gap in auditability of LLM-based scientific agents and provides a structured protocol. However, the evaluation is limited to materials-science tasks without explicit mention of real-world datasets or benchmarks, and the abstract lacks details on comparative baselines or limitations.

Read-first score

Read-first score 54.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 67.

Recency 8%

100

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

Methodology quality 25%

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

Topical relevance 42%

55.8

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

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 68.

Keyword Scores

AI scientist

AI for scientific research

scientific discovery agent

automated scientific discovery

autonomous research agent

automated research

research automation

automated experimentation

experiment design agent

literature review agent

survey generation

paper writing agent

Deep Analysis

Innovations

Hypothesis Evolution Protocol (HEP) that makes hypothesis generation, evaluation, and evolution explicit, auditable operations for LLM agents

Methodology

HEP is a harness that structures LLM agents' scientific reasoning into explicit steps of hypothesis generation, evaluation, and evolution. The approach is evaluated on materials-science research tasks, comparing HEP-equipped agents against planning-style agents.

Key Results

HEP-equipped agents perform the hypothesis-test-evidence-belief cycle, generalize across research questions, and exploit the protocol more fully as the base LLM becomes more capable.

Resources

Paper