Meta与MIT提出ReHarness，重写Agent反思记忆
Meta与MIT提出ReHarness，让Agent在任务结束后根据实际结果回写并修正此前检索到的反思记忆，而非只做增删。该方法用完成的轨迹作为反馈，纠正过度泛化或过时的经验教训。在3个持续学习基准、16个任务和3种backbone上，ReHarness均优于raw-context、memory、reflection和prompt-optimization基线，且经多重比较校正后增益仍显著。
Brilliant paper on agentic continual learning from Meta and MIT.
(bookmark it)
Agent memory harnesses usually only write forward. A lesson goes into the bank and is never checked against whether it helped.
This work proposes ReHarness, which rewrites the reflections an agent retrieved for a task after it sees how that task turned out.
Add and delete operations keep a reflection bank tidy, but they ignore downstream outcomes. ReHarness uses each finished trajectory as feedback on the specific reflections that guided it, so overgeneralized or outdated lessons get corrected.
Across three continual-learning benchmarks with 16 tasks and three backbones, it beats raw-context, memory, reflection, and prompt-optimization baselines, and the gains stay significant after multiple-comparison correction.
arxiv.org/abs/2610.12668
Chat with Paper: academy.dair.ai/papers/recur…
来源：dair_ai · x.com