Compression-Aware Abstention: Teaching LLMs to Refuse When KV-Compression Masks Remove Answer Evidence
感知压缩的弃权:当KV压缩掩码移除答案证据时,教大型语言模型拒绝回答
机构 * University of Southern California(南加州大学)
AI总结 该研究首次将感知压缩的弃权表述为学习问题,通过训练LoRA适配器,在减少LLM因KV缓存压缩产生的幻觉的同时保留正确回答能力,在压缩缓存解码下取得显著提升。
Comments 19 pages, 5 figures. Accepted to the GroundLM workshop at EMNLP 2026. Code, adapters, and datasets: https://github.com/mali-kh/compression-aware-abstention