无需外部信息，分三阶段渐进提升大模型信息抽取能力
ProUIE: A Macro-to-Micro Progressive Learning Method for LLM-based Universal Information Extraction
- 从宏观到微观分三步：先学基础任务，再简化输出格式，最后精细奖励优化
- 36个数据集测试中，小模型超越强基线，尤其在大规模生产场景表现优异
- 适合追求高效、可落地信息抽取的开发者和工程团队
基于大模型的通用信息抽取（UIE）方法通常依赖原始训练数据之外的信息，增加训练复杂度却收益有限。为此，我们提出ProUIE，一种宏到微渐进学习方法，无需引入任何外部信息即可提升UIE性能。该方法包含三个阶段：(i) 宏观级完整建模（CM），在全量训练数据上按内在难度顺序联合学习命名实体识别（NER）、关系抽取（RE）和事件抽取（EE），构建统一抽取基础；(ii) 中观级简化对齐（SA），在采样数据上使用简化目标格式，精简并规范化结构化输出，提升可控性；(iii) 微观级深度探索（DE），通过逐步细粒度奖励（SFR）的GRPO策略，引导模型在结构单元层面进行探索优化。在36个公开数据集上的实验表明，ProUIE持续提升统一抽取效果，在NER和RE上平均优于强指令调优基线，且使用更小的骨干模型，在大规模生产导向的信息抽取中展现明显优势。
原文摘要 · Abstract (English)
LLM-based universal information extraction (UIE) methods often rely on additional information beyond the original training data, which increases training complexity yet often yields limited gains. To address this, we propose ProUIE, a Macro-to-Micro progressive learning approach that improves UIE without introducing any external information. ProUIE consists of three stages: (i) macro-level Complete Modeling (CM), which learns NER, RE, and EE along their intrinsic difficulty order on the full training data to build a unified extraction foundation, (ii) meso-level Streamlined Alignment (SA), which operates on sampled data with simplified target formats, streamlining and regularizing structured outputs to make them more concise and controllable, and (iii) micro-level Deep Exploration (DE), which applies GRPO with stepwise fine-grained rewards (SFR) over structural units to guide exploration and improve performance. Experiments on 36 public datasets show that ProUIE consistently improves unified extraction, outperforming strong instruction-tuned baselines on average for NER and RE while using a smaller backbone, and it further demonstrates clear gains in large-scale production-oriented information extraction.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。