标签
This paper investigates how epistemic stance (qualifiers, attributions) survives memory compression in AI agent memory systems. It finds that making the stance explicit as a labelled field improves retention significantly, while merely lengthening the text does not.
介绍了factwash，一个开源的写入时门控，它通过去除来源归属、模糊限制语或时间语境，确定性地捕捉将传闻变成事实的AI改写。本文分析了何时简单检查就足够、何时需要LLM见证者，并在大型标注语料库上评估了该方法。