arXiv 论文提出智能体自我进化 harness 框架，以智能体自身为优化器
Multi-Task Self-Evolving Harnesses with the Agent as Its Own Optimizer
AI 导读
arXiv 论文（arXiv:2609.38372）提出递归自我改进式框架：同一冻结模型先作为 solver 解题，再作为 proposer 读取完整运行记录并直接编辑运行自身的 harness，进化任务来自五个不同领域基准，训练与 held-out 任务严格分离并额外在五个分布外基准上评估。
推荐理由
论文让同一冻结模型在多任务上自我进化 agent harness，并给出分布外泛化与涌现机制分析，方法思路可迁移。
正文 · 原文
Abstract:A harness is the code around a language-model agent that organizes prompts, calls tools, manages context, and controls execution. As models grow stronger, recent work has begun to let agents improve their own harnesses, a line of work known as self-evolving harnesses. In most existing methods, a separate proposer running on a human-designed harness modifies the solver's harness, and a separate harness is evolved for each benchmark. Real-world tasks come from many domains, so both the evolution and the evaluation of a harness should cover a diverse range of tasks. We propose a framework close to recursive self-improvement: the same frozen model, on the same version of the harness, first solves tasks as the solver and then, as the proposer, reads the complete run records and directly edits the harness that runs it. Each evolution batch draws tasks from five benchmarks in different domains. To measure generalization, training and held-out tasks are strictly separated, and we additionally evaluate on five out-of-distribution benchmarks never used during evolution. We frame the evolution process as deep-learning training with two stages, multi-task pretraining and continual training. Starting from a 49-line seed harness, the harnesses evolved by GLM-5.3 and GPT-5.6 Sol gain about 5 points on the in-distribution benchmarks, matching or exceeding Claude Code and Codex, and the GPT-5.6 Sol harness is on par with Codex on five out-of-distribution benchmarks. Continued evolution on Claw-Eval in the second stage then lifts GPT-5.6 Sol and GLM-5.3-Flash above Codex and Claude Code. We also provide an in-depth analysis of the mechanisms that emerged during evolution, including output truncation, history compaction, completion checks, independent review, and visual inspection. Code and the evolved harnesses are available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2609.38372 [cs.AI] |
| (or arXiv:2609.38372v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.38372 arXiv-issued DOI via DataCite |
Submission history
From: Qiankai Xu [view email]
[v1]
Tue, 29 Sep 2026 18:33:27 UTC (963 KB)
[v2]
Thu, 12 Nov 2026 08:22:19 UTC (1,142 KB)
来源：arXiv cs.AI · arxiv.org