arXiv cs.AI· Junyao Yang, Zongxia Li, Yucheng Shi, Zhongzhi Li, Ruhan Wang, Ruiwen Zhou, Haitao Mi, Shuicheng Yan, Leowei Liang·· 2 天前AI 评分23
Terminal-RSI：通过递归自改写与策略精炼实现递归自我改进的终端智能体
Terminal-RSI: Recursive Self-Improving Terminal Agents via Recursive Self-Rewrite and Policy-Refine
AI 导读
Terminal-RSI 通过交替进行已验证轨迹的递归自改写（RSR）与继承评论家的递归策略精炼（RPR），从 Qwen3.8-27B 基座模型经五阶段训练得到 Terminal-RSI-27B。
正文 · 原文
Abstract:Converting verifier outcomes into useful policy updates and partial progress into complete task resolution is crucial for terminal agents. We present Terminal-RSI, which alternates Recursive Self-Rewrite (RSR) of verified trajectories with critic-inheriting Recursive Policy-Refine (RPR). We formulate Turn-Boundary Skip-Observation GAE (TB-GAE) to retain token-specific baselines while applying decay across interaction turns. Family Dense Reward (FDR) calibrates semantic progress across task families; Success-Prioritized Reward (SPR) separates strict success from bounded partial credit. A Warm-up Critic Model, recursive critic handoff, and stage-specific learning rates coordinate the two refinement phases. The five-stage sequence from Qwen3.8-27B base model through RSR-1, RPR-1, RSR-2, and RPR-2 yields Terminal-RSI-27B: 19.84% on Terminal-Bench 4.0, 86.52% on Terminal-Bench 2.1, 16.95% on Terminal-Bench 3.0, and 79.00% on Terminal-Bench Hard. On LHTB, process reward rises from 0.21 to 0.37 and strict completion from 0/46 to 5/46; SWR100 success doubles from 3.00% to 6.00%. By recursively coupling verified trajectory rewriting with critic-inheriting policy refinement, each retained stage consolidates discovered behaviors and progressively advances the agent from partial local repairs to strict terminal verification.
| Comments: | 28 pages, 16 figures, 5 tables |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.22192 [cs.AI] |
| (or arXiv:2610.22192v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22192 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Junyao Yang [view email]
[v1]
Thu, 15 Oct 2026 16:43:56 UTC (11,576 KB)
来源：arXiv cs.AI · arxiv.org