AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage

TL;DR

AutoReproduce is a multi-agent framework that automates the reproduction of AI research experiments by leveraging paper lineage and a new benchmark, significantly improving fidelity and execution performance.

Contribution

It introduces a novel paper lineage method and an autonomous framework for reproducing experiments, along with a benchmark for evaluating reproduction fidelity.

Findings

AutoReproduce outperforms existing baselines across all metrics.

It significantly improves reproduction fidelity and execution performance.

The framework is validated on PaperBench and PaperReproduce.

Abstract

Efficient reproduction of research papers is pivotal to accelerating scientific progress. However, the increasing complexity of proposed methods often renders reproduction a labor-intensive endeavor, necessitating profound domain expertise. To address this, we introduce the paper lineage, which systematically mines implicit knowledge from the cited literature. This algorithm serves as the backbone of our proposed \ours, a multi-agent framework designed to autonomously reproduce experimental code in a complete, end-to-end manner. To ensure code executability, \ours incorporates a sampling-based unit testing strategy for rapid validation. To assess reproduction capabilities, we introduce \ourbench, a benchmark featuring verified implementations, alongside comprehensive metrics for evaluating both reproduction and execution fidelity. Extensive evaluations on PaperBench and \ourbench demonstrate that \ours consistently surpasses existing baselines across all metrics. Notably, it yields substantial improvements in reproduction fidelity and final execution performance. The code is available at https://github.com/AI9Stars/AutoReproduce.