ClawLoop: Self-Improving Agents via Harness-Native Training Loop

The rapid advancement of Large Language Models (LLMs) has driven their deployment toward general-purpose harness environments, where LLM-based agents operate in sandbox workspaces and interact directly with file systems and shells.

大型语言模型（LLM）的快速发展推动了其向通用测试环境（harness environments）的部署，在这种环境中，基于 LLM 的智能体在沙盒工作区中运行，并直接与文件系统和 Shell 进行交互。

While numerous post-training pipelines exist for tool-use agents, their counterparts for harness-based agents remain scarce and are predominantly one-shot and teacher-dependent: a stronger external model synthesizes tasks and collects trajectories, the target agent is trained once, and evaluation results are consumed by human researchers rather than fed back into the loop.

虽然针对工具使用智能体的训练后流水线已有很多，但针对基于测试环境的智能体的同类方案仍然稀缺，且主要是一次性的且依赖于教师模型：即由更强大的外部模型合成任务并收集轨迹，目标智能体仅接受一次训练，且评估结果由人类研究人员消费，而非反馈回循环中。

Such pipelines are inherently bounded by the capability of the external model and the scope of human prior, leaving no room for self-improvement. To address these limitations, we propose ClawLoop, a framework that enables self-improving without any external teacher model and with minimal prior dependencies.

此类流水线在本质上受到外部模型能力和人类先验范围的限制，没有自我提升的空间。为了解决这些局限性，我们提出了 ClawLoop，这是一个无需任何外部教师模型且仅需极少先验依赖即可实现自我提升的框架。

ClawLoop composes three phases into a self-sustaining cycle: Task Synthesis, which generates a diverse pool of training tasks via two complementary construction routines, applies diagnosis-conditioned difficulty growth followed by both agent-free and agent-based task validation; Training, which updates the agent via SFT and/or RL; and Diagnosis and Feedback, which analyzes the agent’s rollout behaviors to produce a capability analysis report that will be used as feedback for task synthesis in the next iteration.

ClawLoop 将三个阶段组成一个自我维持的循环：任务合成（Task Synthesis），通过两种互补的构建程序生成多样化的训练任务池，应用基于诊断的难度增长，并进行无智能体和基于智能体的任务验证；训练（Training），通过 SFT 和/或 RL 更新智能体；以及诊断与反馈（Diagnosis and Feedback），分析智能体的执行行为以生成能力分析报告，该报告将作为下一轮任务合成的反馈。

Using Qwen3.6-27B as the loop model, ClawLoop yields consistent improvements on SkillsBench, Claw-Eval-Live, and ClawGym-Bench. The iteratively refined tasks are further transferred to smaller-scale Qwen3.5-4B and 9B models with gains across all benchmarks, validating data quality independently of the loop model.

使用 Qwen3.6-27B 作为循环模型，ClawLoop 在 SkillsBench、Claw-Eval-Live 和 ClawGym-Bench 上取得了持续的改进。经过迭代优化的任务被进一步迁移到较小规模的 Qwen3.5-4B 和 9B 模型中，并在所有基准测试中均获得提升，验证了数据质量独立于循环模型。