发表机构
New York University（纽约大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对长期智能体记忆压缩导致的分布式证据悖论，提出DerivAudit审计框架，通过证据范围、组合有效性和准入可靠性三个要求，验证记忆是否由写入时历史支持，实验恢复近60%看似无支持记忆的支持，但扩展证据不能保证准入可靠。
AI 中文摘要
长期运行的LLM智能体将过去的交互压缩为持久记忆，这些记忆可能被复用为后续任务的前提。这产生了一个独特的推导问题：记忆是否确实从交互历史所支持的内容中推导而来。相关证据可能分散在早期的交互中，而压缩可能引入历史从未确立的关系或事件状态。因此，一条有效的记忆可能看似缺乏支持，因为其引用遗漏了相关证据，而个别得到支持的事实可能被组合成一个历史从未确立的更强陈述。我们通过三个耦合的要求来刻画这一问题：（1）证据范围；（2）组合有效性；（3）准入可靠性。因此，我们询问写入时可用的交互历史是否支持进入持久记忆的内容。我们引入了DerivAudit，一个用于审计记忆是否确实由写入时可用的历史支持的框架。该审计区分了三个问题：支持性证据是否超出写入者提供的引用范围，组合后的记忆是否引入了无支持的含义，以及写入时的准入决策如何影响后续的记忆使用。在两个自然记忆语料库上，使用更广泛的写入前历史进行审计，为近60%的仅从引用看似无支持的记忆恢复了支持，而17-21%在扩展后仍无支持。然而，更广泛的证据本身并不能使准入可靠：无支持的记忆在验证模型中仍经常被准入，且仅扩展证据在两个骨干模型上使其恶化。
英文摘要
Long-running LLM agents compress past interactions into persistent memories that may be reused as premises for later tasks. This creates a distinct derivation problem: whether the memory actually follows from what the interaction history supports. Relevant evidence may be scattered across earlier interactions, while compression can introduce relations or event status that the history never established. A valid memory may therefore appear unsupported because its citations omit relevant evidence, while individually supported facts may be composed into a stronger statement the history never established. We characterize this problem through three coupled requirements: (1) Evidence scope; (2) Compositional validity; (3) Admission reliability. We therefore ask whether the interaction history available at write time supports what enters persistent memory. We introduce DerivAudit, a framework for auditing whether a memory is actually supported by the history available when it was written. The audit separates three questions: whether supporting evidence lies beyond writer-provided citations, whether the composed memory introduces unsupported meaning, and how write-time admission decisions affect later memory use. Across two natural memory corpora, audits using broader pre-write history recover support for nearly 60% of memories that appear unsupported from citations alone, while 17-21% remain unsupported after expansion. Yet broader evidence does not by itself make admission reliable: unsupported memories are still frequently admitted across verification models, and evidence expansion alone worsens it on two backbones.
Comments21 pages,9 tables, 5 figures
