SRJudge: Empowering Large Language Models with Selective Reasoning for Fine-Grained Knowledge Concept Tagging
SRJudge:赋予大语言模型选择性推理能力以实现细粒度知识概念标注
机构 * Guangdong Institute of Smart Education, Jinan University(暨南大学广东智慧教育研究院) ; School of Physical Education, Jinan University(暨南大学体育学院) ; Guangdong Provincial Key Laboratory of Speed Capability Research(广东省速度能力研究重点实验室) ; Sapient Intelligence Pte Ltd
AI总结 本文提出SRJudge三阶段框架,通过选择器缩小候选、推理器精细推理、评判器评估,增强大语言模型在细粒度知识概念标注中的选择性推理能力,并在生物和物理数据集上超越现有基线。
Comments Accepted by IJCAI 2026