Recuris:递归经验-工作记忆演化提升长时程智能体性能
原标题:Paper page - Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses
AI 摘要
Recuris 提出了一种递归记忆架构,通过工作记忆跟踪任务进度并指导技能选择,利用经验记忆和验证门控的局部更新,提升长时程智能体任务的成功率。在四个基准和十个模型上的测试中,Recuris 在 37 个模型-基准组合中的 35 个上提升了任务成功率,例如在 tau-bench 上为 GPT-5.6 Sol 增加 17.8 分,为 Claude Opus 5 增加 15.6 分,使其达到 87.9%。该架构在更长交互任务中优势更明显,最长任务上提升 32.2 分,常见长时程失败减少 80%。
正文节选
Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses Abstract Recuris introduces a recursive memory architecture that tracks progress and guides skill selection to improve long-horizon agent success through localized, validation-gated updates. Recursive self-improvement (RSI) remains hard in long-horizon tasks, where growing histories obscure the task state and misalign skill invocation. We introduce Recuris, a recursive Experiential-Working Memory architecture for lo