LLM training improves actionable peer review feedback for paper revisions
ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation
Computation and Language
Summary
Getting useful feedback that tells authors how to improve their papers before submission is hard. The authors study how to make AI-generated peer reviews that not only spot problems but also suggest clear fixes. They use author rebuttals to guide the AI in producing review comments tied to parts of the paper and real revision ideas. Their method, ActReview, shows better helpfulness and grounding than past approaches while still needing to improve on technical accuracy. They also created a new dataset and evaluation benchmark to measure progress.
What this means in practice
- •For academic journal editors: Automate producing detailed and actionable peer reviews that help authors improve paper drafts before official submission.
- •For conference program committees: Generate grounded reviewer feedback linked to specific paper weaknesses and revision plans to aid decision making and author guidance.
Authors
Yiling Ma, Yilun Zhao, Sihong Wu, Ziyu Chen, Manasi Patwardhan, Arman Cohan
Abstract
As LLMs are increasingly used for pre-submission self-review, there is growing demand for feedback that not only identifies weaknesses but also guides authors toward concrete revisions. We study this as Actionable Peer-review Generation and decompose it into two subtasks: diagnostic claim generation and revision suggestion generation. We introduce ActReview, a rebuttal-guided post-training framework that connects paper-specific diagnoses to concrete, grounded revision plans. Our central insight is that author rebuttals reveal plausible actions for addressing reviewer concerns and can therefore provide latent supervision for revision-oriented feedback. From real review-rebuttal threads on OpenReview, we construct ActReview-40K by aligning reviewer weaknesses with author responses and grounding the resulting feedback in localized paper evidence. We post-train Qwen3-8B-Base with multi-task supervised fine-tuning followed by GRPO using candidate-aware, weakness-specific rubric rewards. We also introduce ActReview-Bench, a human-curated benchmark of 1,000 instances for evaluating diagnostic quality and revision usefulness. Experiments show that ActReview outperforms prior specialized review-generation models on actionability and grounding while remaining competitive with strong prompt-based LLMs. Human evaluation confirms improved revision usefulness while revealing a remaining gap in technical accuracy, and additional analyses support generalization to held-out papers and robustness across independent judges.