Reproduce “Chain-of-Factors Paper-Reviewer Matching” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yu Zhang, Yanzhen Shen, SeongKu Kang, Xiusi Chen, Bowen Jin, Jiawei Han
Researchers from Texas A&M University and the University of Illinois Urbana-Champaign developed the Chain-of-Factors (CoF) framework for paper-reviewer matching, which unifies semantic, topic, and citation factors through an instruction-guided language model and a sequential coarse-to-fine inference process. This approach consistently outperformed classical and scientific pre-trained language model baselines across four datasets and improved performance on individual factor tasks.
Key insights
- Jointly considering semantic, topic, and citation factors within a unified framework provides a more comprehensive and accurate assessment of expertise relevance than single-factor approaches.
- Instruction-guided pre-training allows a single, shared language model to learn distinct, factor-specific representations while capturing common knowledge.
- A coarse-to-fine chained inference strategy, which progressively filters candidate reviewers based on different relevance factors, enhances matching accuracy over single-step aggregation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.