長時程智能體駕馭框架之遞迴經驗-工作記憶演化

摘要

遞迴自我改進（RSI）在長時程任務中仍然困難，因為不斷增長的歷史會模糊任務狀態，並使技能調用失準。我們提出 Recuris，一種用於長時程代理框架的遞迴式經驗-工作記憶架構，其中工作記憶追蹤任務進度，並引導從經驗記憶中選擇技能，使技能的使用立足於當前需求，而非完整的歷史。這種耦合也將執行轉化為結構化證據，可將失敗定位到特定的記憶組件。在任務之間，一個固定的後設代理將這些證據轉化為對技能記憶的局部化、經驗證閘控的更新，這些更新重塑執行並產生新證據，形成一個有界的遞迴記憶演化迴圈。在四個長時程基準測試與十個模型上，Recuris 在 37 個已完成的模型-基準配對中，提升了 35 個的任務成功率，將前沿模型推向 SOTA 等級的任務成功率：在 tau-bench 上，它為 GPT-5.6 Sol 增加 +17.8 個百分點，為 Claude Opus 5 增加 +15.6 個百分點，使 Opus 5 達到 87.9%；在 SkillFlow 上，為 Qwen3.6-27B/35B 分別增加 +16.6/+13.5 個百分點。隨著互動時程的增長，優勢會擴大，在最長任務上達到 +32.2 個百分點，而常見的長時程失敗最多減少 80%。這些結果將遞迴演化記憶定位為 RSI 的可擴展基礎，使代理能夠持續將累積的經驗轉化為越來越有效的長時程行為。程式碼：https://github.com/Gen-Verse/Recuris

English

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 long-horizon agent harnesses, in which Working Memory tracks task progress and guides skill selection from Experiential Memory, grounding skill use in current needs rather than the full history. This coupling also turns execution into structured evidence that localizes failures to specific memory components. Across tasks, a fixed Meta-Agent turns that evidence into localized, validation-gated updates to Skill Memory that reshape execution and yield new evidence, forming a bounded recursive memory-evolution loop. Across four long-horizon benchmarks and ten models, Recuris improves task success in 35 of the 37 completed model-benchmark pairs, carrying frontier models to SOTA-level task success: on tau-bench it adds +17.8 points to GPT-5.6 Sol and +15.6 to Claude Opus 5, taking Opus 5 to 87.9%, and +16.6/+13.5 points on Qwen3.6-27B/35B on SkillFlow. The advantage widens as the interaction horizon grows, to +32.2 points on the longest tasks, and common long-horizon failures fall by up to 80%. These results position recursively evolving memory as a scalable foundation for RSI, enabling agents to continuously transform accumulated experience into increasingly effective long-horizon behavior. Code: https://github.com/Gen-Verse/Recuris