DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Process

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhu, Minjun, Weng, Yixuan, Yang, Linyi, Zhang, Yue
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915191660216320
author Zhu, Minjun
Weng, Yixuan
Yang, Linyi
Zhang, Yue
author_facet Zhu, Minjun
Weng, Yixuan
Yang, Linyi
Zhang, Yue
contents Large Language Models (LLMs) are increasingly utilized in scientific research assessment, particularly in automated paper review. However, existing LLM-based review systems face significant challenges, including limited domain expertise, hallucinated reasoning, and a lack of structured evaluation. To address these limitations, we introduce DeepReview, a multi-stage framework designed to emulate expert reviewers by incorporating structured analysis, literature retrieval, and evidence-based argumentation. Using DeepReview-13K, a curated dataset with structured annotations, we train DeepReviewer-14B, which outperforms CycleReviewer-70B with fewer tokens. In its best mode, DeepReviewer-14B achieves win rates of 88.21\% and 80.20\% against GPT-o1 and DeepSeek-R1 in evaluations. Our work sets a new benchmark for LLM-based paper review, with all resources publicly available. The code, model, dataset and demo have be released in http://ai-researcher.net.
format Preprint
id arxiv_https___arxiv_org_abs_2503_08569
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Process
Zhu, Minjun
Weng, Yixuan
Yang, Linyi
Zhang, Yue
Computation and Language
Machine Learning
Large Language Models (LLMs) are increasingly utilized in scientific research assessment, particularly in automated paper review. However, existing LLM-based review systems face significant challenges, including limited domain expertise, hallucinated reasoning, and a lack of structured evaluation. To address these limitations, we introduce DeepReview, a multi-stage framework designed to emulate expert reviewers by incorporating structured analysis, literature retrieval, and evidence-based argumentation. Using DeepReview-13K, a curated dataset with structured annotations, we train DeepReviewer-14B, which outperforms CycleReviewer-70B with fewer tokens. In its best mode, DeepReviewer-14B achieves win rates of 88.21\% and 80.20\% against GPT-o1 and DeepSeek-R1 in evaluations. Our work sets a new benchmark for LLM-based paper review, with all resources publicly available. The code, model, dataset and demo have be released in http://ai-researcher.net.
title DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Process
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2503.08569