Salvato in:
Dettagli Bibliografici
Autori principali: Zhao, Xuanle, Sang, Zilin, Li, Yuxuan, Shi, Qi, Zhao, Weilun, Wang, Shuo, Zhang, Duzhen, Han, Xu, Liu, Zhiyuan, Sun, Maosong
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2505.20662
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910161101127680
author Zhao, Xuanle
Sang, Zilin
Li, Yuxuan
Shi, Qi
Zhao, Weilun
Wang, Shuo
Zhang, Duzhen
Han, Xu
Liu, Zhiyuan
Sun, Maosong
author_facet Zhao, Xuanle
Sang, Zilin
Li, Yuxuan
Shi, Qi
Zhao, Weilun
Wang, Shuo
Zhang, Duzhen
Han, Xu
Liu, Zhiyuan
Sun, Maosong
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20662
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage
Zhao, Xuanle
Sang, Zilin
Li, Yuxuan
Shi, Qi
Zhao, Weilun
Wang, Shuo
Zhang, Duzhen
Han, Xu
Liu, Zhiyuan
Sun, Maosong
Artificial Intelligence
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.
title AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage
topic Artificial Intelligence
url https://arxiv.org/abs/2505.20662