From Similarity to Structure: Training-free LLM Context Compression with Hybrid Graph Priors
从相似性到结构:基于混合图先验的训练自由LLM上下文压缩
机构 * School of Computer Science and Engineering, University of Electronic Science and Technology of China(电子科技大学计算机科学与工程学院) ; School of Information and Software Engineering, University of Electronic Science and Technology of China(电子科技大学信息与软件工程学院) ; Department of Computer Science and Engineering, Kyung Hee University(Kyung Hee大学计算机科学与工程系) ; Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University(香港理工大学工业与系统工程系) ; School of Robotics and Advanced Manufacture, Harbin Institute of Technology, Shenzhen(哈尔滨工业大学深圳校区机器人与先进制造学院)
专题命中 长文档RAG :dense retrieval(abstract);分类 cs.CL、cs.AI
AI总结 本文提出一种无需训练的上下文压缩框架,利用混合图先验选择句子,通过构建稀疏句子图、提取主题骨架并结合可解释评分进行冗余抑制,实验证明在长文档基准上表现更优。