---
title: "递归经验工作记忆演化：面向长程智能体平台的 Recuris 架构"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.24876v1"
permalink: "https://feed.astralor.com/items/itm_15868b7c171bd6b2d60e033f"
published: "2026-08-25T17:56:35.000Z"
score: 75
channel: "tech"
tags: ["Recuris","元智能体","工作记忆","递归自我改进","长程智能体"]
---

# 递归经验工作记忆演化：面向长程智能体平台的 Recuris 架构

## AI 点评

> 让我感兴趣的不是又刷了几个点，而是 Recuris 把记忆演化变成有界、可校验的闭环，并且能将失败定位到具体记忆组件。对做长程 Agent 框架或 RSI 研究的团队来说，值得立刻读原文并试跑代码。

## 速览

论文提出 Recuris，一种递归经验工作记忆架构，用于长程 Agent 任务中解耦任务历史与技能调用。工作记忆追踪任务进度并引导经验记忆中的技能选择，执行过程成为结构化证据，由固定元智能体生成经校验的局部技能更新，形成有界递归记忆演化循环。在四个基准和十个模型上，Recuris 在 37 个模型基准组合中的 35 个上提升成功率，并将 Claude Opus 5 在 tau-bench 上带到 87.9%，最长任务上提升达 32.2 个百分点。

## 关键点

- 提出 Recuris，一种用于长程智能体的递归经验工作记忆架构。
- 通过固定元智能体将执行证据转化为可校验的局部技能更新，形成有界递归演化闭环。
- 在 tau-bench 与 SkillFlow 上刷新前沿模型表现，如 Claude Opus 5 达 87.9%。
- 优势随任务长度扩大，最长任务最高提升 32.2 个百分点，常见失败下降最高 80%。

## 全文

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

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_15868b7c171bd6b2d60e033f