The Year in AI Papers
The year/Independent research

Paper 2603.17187

MetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild

Peng Xia, Jianwen Chen, Xinyu Yang, Haoqin Tu, Jiaqi Liu, Kaiwen Xiong, Siwei Han, Shi Qiu, Haonian Ji, Yuyin Zhou, Zeyu Zheng, Cihang Xie, et al.

Published
Mar 2026
Research lab
Independent
Citations
20
GitHub
3.5K stars

01 In brief

Summary

MetaClaw is a continual meta-learning framework that enables deployed LLM agents to evolve through two complementary mechanisms: skill-driven fast adaptation and opportunistic policy optimization.

Skill-driven fast adaptation analyzes failure trajectories to synthesize new behavioral instructions, which are immediately injected into the agent's prompt without downtime.

Opportunistic policy optimization uses reinforcement learning with a process reward model to update model weights via cloud LoRA fine-tuning, triggered only during user-inactive windows by the Opportunistic Meta-Learning Scheduler (OMLS).

A skill generation versioning mechanism ensures that policy optimization trains only on post-adaptation trajectories, preventing stale reward contamination.

Experiments on MetaClaw-Bench (934 questions, 44 simulated workdays) show that skill-driven adaptation improves accuracy by up to 32.2% relative, and the full pipeline advances Kimi-K2.5 from 21.4% to 40.6% accuracy, nearly closing the gap with GPT-5.2's baseline (41.1%).

On AutoResearchClaw, a 23-stage autonomous research pipeline, skill injection alone improves composite robustness by 18.3%.

MetaClaw requires no local GPU and scales to production-size LLMs via a proxy-based architecture, enabling continuous improvement without service interruption.

02 From the paper

Abstract

Large language model (LLM) agents are increasingly used for complex tasks, yet deployed agents often remain static, failing to adapt as user needs evolve. This creates a tension between the need for continuous service and the necessity of updating capabilities to match shifting task distributions. On platforms like OpenClaw, which handle diverse workloads across 20+ channels, existing methods either store raw trajectories without distilling knowledge, maintain static skill libraries, or require disruptive downtime for retraining. We present MetaClaw, a continual meta-learning framework that jointly evolves a base LLM policy and a library of reusable behavioral skills. MetaClaw employs two complementary mechanisms. Skill-driven fast adaptation analyzes failure trajectories via an LLM evolver to synthesize new skills, enabling immediate improvement with zero downtime. Opportunistic policy optimization performs gradient-based updates via cloud LoRA fine-tuning and Reinforcement Learning with a Process Reward Model (RL-PRM). This is triggered during user-inactive windows by the Opportunistic Meta-Learning Scheduler (OMLS), which monitors system inactivity and calendar data. These mechanisms are mutually reinforcing: a refined policy generates better trajectories for skill synthesis, while richer skills provide higher-quality data for policy optimization. To prevent data contamination, a versioning mechanism separates support and query data. Built on a proxy-based architecture, MetaClaw scales to production-size LLMs without local GPUs. Experiments on MetaClaw-Bench and AutoResearchClaw show that skill-driven adaptation improves accuracy by up to 32% relative. The full pipeline advances Kimi-K2.5 accuracy from 21.4% to 40.6% and increases composite robustness by 18.3%. Code is available at https://github.com/aiming-lab/MetaClaw.