Refined reproducibility tools improve success in machine learning experiments
ReproBridge: Building Reproducibility into ML Research Artifacts
Artificial Intelligence
Summary
It can be hard to copy how machine learning experiments are done because the details are scattered in papers and code. The authors studied whether saving clear, organized instructions and setup helps others repeat the experiments more easily. They made a system called ReproBridge that builds these clear and complete experiment guides. Using these guides, different AI models reproduced experiments more accurately and with less cost. This shows that keeping good records of how experiments run helps anyone replicate them better.
What this means in practice
- •For machine learning engineers: Use refined reproducibility guides to more reliably replicate published machine learning experiments and reduce costs of experimentation.
- •For software quality assurance teams: Implement standardized artifact designs to improve reproducibility checks and auditing in AI software testing pipelines.
Authors
Yubo Wang, Rui Meng, Jaehyun Nam, Tomas Pfister, Jinsung Yoon
Abstract
Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, released code, and supporting artifacts. For coding agents, reproduction success can therefore depend on their ability to fill in missing details, rather than on the released artifacts alone. We ask whether preserving recovered procedures and decisions improves subsequent independent reproduction. We define a standardized artifact design comprising a refined codebase, environment setup, experiment launchers, and a reproduction guide, and introduce ReproBridge, an agent workflow for constructing these refined reproducible artifacts. ReproBridge-Bench provides 1,029 tasks from 387 experiments across 36 papers, anchored by five human-verified reference scores per task. Agents attempt tasks in fresh sessions with raw code or refined artifacts; a separate Auditor checks execution evidence. Across four evaluated models, refined artifacts raise mean reproduction success from 76.4% to 92.6% and reduce downstream API spending by 19.7% on average. The cross-model success-rate spread narrows from 21.6 to 3.9 percentage points, with all four models achieving mean success rates above 90%. Removing any component reduces success in the ablation study, with the largest losses from code and environment removal. These results suggest that preserving execution knowledge helps even models with lower raw-code success reproduce experiments more reliably and at lower API cost.