Reproduce “RUIE: Retrieval-based Unified Information Extraction using Large Language Model” 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
Xincheng Liao, Junwen Duan, Yixi Huang, Jianxin Wang
RUIE trains a smaller retriever to select labeled examples from mixed-task pools for a frozen large language model for information extraction. It uses language-model preferences and a keyword-enhanced reward model to train the retriever. On held-out relation extraction, its average span-based Micro-F1 is 51.55, versus 47.50 for BM25 with eight demonstrations; its held-out event-argument score is slightly lower than BM25’s.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.