The Story is Not the Science: Execution-Grounded Evaluation of Mechanistic Interpretability Research
- URL: http://arxiv.org/abs/2602.18458v1
- Date: Thu, 05 Feb 2026 19:00:02 GMT
- Title: The Story is Not the Science: Execution-Grounded Evaluation of Mechanistic Interpretability Research
- Authors: Xiaoyan Bai, Alexander Baumgartner, Haojia Sun, Ari Holtzman, Chenhao Tan,
- Abstract summary: We address the challenges of scalability and rigor by flipping the dynamic and developing AI agents as research evaluators.<n>We use mechanistic interpretability research as a testbed, build standardized research output, and develop MechEvalAgent.<n>Our work demonstrates the potential of AI agents to transform research evaluation and pave the way for rigorous scientific practices.
- Score: 56.80927148740585
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Reproducibility crises across sciences highlight the limitations of the paper-centric review system in assessing the rigor and reproducibility of research. AI agents that autonomously design and generate large volumes of research outputs exacerbate these challenges. In this work, we address the growing challenges of scalability and rigor by flipping the dynamic and developing AI agents as research evaluators. We propose the first execution-grounded evaluation framework that verifies research beyond narrative review by examining code and data alongside the paper. We use mechanistic interpretability research as a testbed, build standardized research output, and develop MechEvalAgent, an automated evaluation framework that assesses the coherence of the experimental process, the reproducibility of results, and the generalizability of findings. We show that our framework achieves above 80% agreement with human judges, identifies substantial methodological problems, and surfaces 51 additional issues that human reviewers miss. Our work demonstrates the potential of AI agents to transform research evaluation and pave the way for rigorous scientific practices.
Related papers
- FIRE-Bench: Evaluating Agents on the Rediscovery of Scientific Insights [63.32178443510396]
We introduce FIRE-Bench (Full-cycle Insight Rediscovery Evaluation), a benchmark that evaluates agents through the rediscovery of established findings.<n>Even the strongest agents achieve limited rediscovery success (50 F1), exhibit high variance across runs, and display recurring failure modes in experimental design, execution, and evidence-based reasoning.
arXiv Detail & Related papers (2026-02-02T23:21:13Z) - Towards AI Agents Supported Research Problem Formulation [3.6732711233211663]
Poorly formulated research problems can compromise the practical relevance of Software Engineering studies.<n>This vision paper explores the use of artificial intelligence agents to support SE researchers during the early stage of a research project.
arXiv Detail & Related papers (2025-12-14T14:44:27Z) - AstaBench: Rigorous Benchmarking of AI Agents with a Scientific Research Suite [75.58737079136942]
We present AstaBench, a suite that provides the first holistic measure of agentic ability to perform scientific research.<n>Our suite comes with the first scientific research environment with production-grade search tools.<n>Our evaluation of 57 agents across 22 agent classes reveals several interesting findings.
arXiv Detail & Related papers (2025-10-24T17:10:26Z) - ReviewerToo: Should AI Join The Program Committee? A Look At The Future of Peer Review [23.630458187587223]
ReviewerToo is a framework for studying and deploying AI-assisted peer review.<n>It supports systematic experiments with specialized reviewer personas and structured evaluation criteria.<n>We show how AI can enhance consistency, coverage, and fairness while leaving complex evaluative judgments to domain experts.
arXiv Detail & Related papers (2025-10-09T23:53:19Z) - The More You Automate, the Less You See: Hidden Pitfalls of AI Scientist Systems [11.543423308064275]
AI scientist systems are capable of executing the full research workflow from hypothesis generation to paper writing.<n>This lack of scrutiny poses a risk of introducing flaws that could undermine the integrity, reliability, and trustworthiness of their research outputs.<n>We identify four potential failure modes in contemporary AI scientist systems: inappropriate benchmark selection, data leakage, metric misuse, and post-hoc selection bias.
arXiv Detail & Related papers (2025-09-10T16:04:24Z) - Automatic Reviewers Fail to Detect Faulty Reasoning in Research Papers: A New Counterfactual Evaluation Framework [55.078301794183496]
We focus on a core reviewing skill that underpins high-quality peer review: detecting faulty research logic.<n>This involves evaluating the internal consistency between a paper's results, interpretations, and claims.<n>We present a fully automated counterfactual evaluation framework that isolates and tests this skill under controlled conditions.
arXiv Detail & Related papers (2025-08-29T08:48:00Z) - The AI Imperative: Scaling High-Quality Peer Review in Machine Learning [49.87236114682497]
We argue that AI-assisted peer review must become an urgent research and infrastructure priority.<n>We propose specific roles for AI in enhancing factual verification, guiding reviewer performance, assisting authors in quality improvement, and supporting ACs in decision-making.
arXiv Detail & Related papers (2025-06-09T18:37:14Z) - ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Models [56.08917291606421]
ResearchAgent is an AI-based system for ideation and operationalization of novel work.<n>ResearchAgent automatically defines novel problems, proposes methods and designs experiments, while iteratively refining them.<n>We experimentally validate our ResearchAgent on scientific publications across multiple disciplines.
arXiv Detail & Related papers (2024-04-11T13:36:29Z)
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.