AutoLab-SAI-SJTU/Paper2Rebuttal

What it solves

RebuttalAgent helps academic researchers handle the stressful and complex process of writing rebuttals for peer-reviewed papers. It automates the analysis of reviewer comments and the search for supporting evidence, reducing the manual effort required to address critiques and draft formal responses.

How it works

The system employs a multi-agent pipeline that processes a paper PDF and reviewer comments through several specialized stages:

1. Parsing & Extraction: It uses Docling to convert PDFs to text and extracts specific issues from reviewer comments, assigning them priority levels.
2. Evidence Gathering: It automatically searches arXiv for relevant literature to support the author's arguments.
3. Strategy Development: Agents generate data-driven rebuttal strategies and to-do lists, which are then reviewed and refined.
4. Human-in-the-Loop: Authors can provide natural language feedback to iteratively improve the strategies before the final draft is written.
5. Final Drafting: A dedicated writer and reviewer agent produce a polished, conference-ready rebuttal letter.

Who it’s for

Academic researchers and students who are submitting papers to conferences (such as ACL) and need a structured, AI-assisted way to respond to reviewer feedback.

Highlights

• Multi-Agent Workflow: Uses a chain of specialized agents for encoding, retrieval, strategy generation, and polishing.
• Automated Literature Search: Integrates with arXiv to find supporting papers automatically.
• Interactive Refinement: Allows authors to guide the AI through a feedback loop to ensure strategies are accurate.
• PDF Integration: Built-in support for parsing complex academic PDFs via Docling.