上下文压缩中的约束丢失：评估与解决方案
原标题：Lost in Compaction: Evaluating Side-Constraint Loss under Context Compaction
AI 摘要
宾夕法尼亚州立大学的研究团队发现，在上下文压缩过程中，用户发出的会话约束（如“未经确认不要删除邮件”）会被静默丢弃，导致AI系统违反用户指令。他们提出了CompInt评估套件，涵盖多轮对话、智能体轨迹和长周期研究三种场景，并测试了多种压缩器，结果显示平均仅保留17%的约束。团队还开发了一种与压缩器并行的SC感知提取器，在不修改压缩器或LLM的情况下，将约束保留率提升至90%以上。
正文节选
Lost in Compaction: Evaluating Side-Constraint Loss under Context Compaction Abstract When the context window is under pressure, LLM systems compact prior context to continue ongoing tasks. We identify a class of user-issued instructions, Session Constraints (SCs), such as "do not delete any emails until I confirm," that are meant to constrain LLM’s behavior for the remainder of a session but are silently dropped during compaction. To quantify this loss, we introduce CompInt, an evaluation suite