Beyond path selection: Better LLMs for Scientific Information Extraction with MimicSFT and Relevance and Rule-induced(R$^2$)GRPO

TL;DR

This paper introduces MimicSFT and R$^2$GRPO, two novel training methods that significantly enhance the reasoning and memorization capabilities of large language models for scientific information extraction tasks.

Contribution

The paper proposes a two-stage training framework combining MimicSFT and R$^2$GRPO, which improves LLMs' reasoning capacity without requiring high-quality chain-of-thought data.

Findings

R$^2$GRPO with MimicSFT outperforms baseline LLMs in relation extraction.

Both methods improve reasoning capacity in scientific IE benchmarks.

The approach surpasses specialized supervised models in key tasks.

Abstract

Previous study suggest that powerful Large Language Models (LLMs) trained with Reinforcement Learning with Verifiable Rewards (RLVR) only refines reasoning path without improving the reasoning capacity in math tasks while supervised-finetuning(SFT) with distillation can. We study this from the view of Scientific information extraction (SciIE) where LLMs and reasoning LLMs underperforms small Bert-based models. SciIE require both the reasoning and memorization. We argue that both SFT and RLVR can refine the reasoning path and improve reasoning capacity in a simple way based on SciIE. We propose two-stage training with 1. MimicSFT, using structured reasoning templates without needing high-quality chain-of-thought data, 2. R$^2$GRPO with relevance and rule-induced rewards. Experiments on scientific IE benchmarks show that both methods can improve the reasoning capacity. R$^2$GRPO with mimicSFT surpasses baseline LLMs and specialized supervised models in relation extraction. Our code is available at https://github.com/ranlislz/R2GRPO.