PEARL: Auditable Repair for Scientific Reasoning Graph Extraction
PEARL:科学推理图提取的可审计修复
机构 * School of Computer Science, Wuhan University(武汉大学计算机科学学院) ; Artificial Intelligence Innovation and Incubation, Fudan University(复旦大学人工智能创新与孵化) ; Department of Computer Science, Faculty of Science, University of Bath(巴斯大学理学院计算机科学系) ; College of Computer Science and Artificial Intelligence, Fudan University(复旦大学计算机科学与人工智能学院)
专题命中 领域大模型 :LLM(abstract,abstract_cn);分类 cs.AI
AI总结 针对科学推理图提取中LLMs输出有问题的情况,提出无需训练的PEARL框架,通过特定模式和反馈修复推理图。在ARCHE基准测试中,该框架提升了严格通过率和平均REA,为相关工作流程提供可靠性层。
Comments Accepted at WAICA 2026 Multi-Modal Agents for Science Workshop