Training LLM Agents for Spontaneous, Reward-Free Self-Evolution via World Knowledge Exploration

TL;DR

This paper introduces a method for training large language model agents to spontaneously self-evolve and adapt to unseen environments without external rewards, by learning to explore and summarize world knowledge.

Contribution

It presents an outcome-based reward mechanism for intrinsic self-evolution training, enabling agents to adapt independently at inference time.

Findings

20% performance increase on WebVoyager and WebWalker tasks.

Generated world knowledge allows a smaller model to outperform larger unassisted models.

Agents can self-evolve without external rewards or human instructions.

Abstract

Most agents today ``self-evolve'' by following rewards and rules defined by humans. However, this process remains fundamentally dependent on external supervision; without human guidance, the evolution stops. In this work, we train agents to possess an intrinsic meta-evolution capability to spontaneously learn about unseen environments prior to task execution. To instill this ability, we design an outcome-based reward mechanism that measures how much an agent's self-generated world knowledge improves its success rate on downstream tasks. This reward signal is used exclusively during the training phase to teach the model how to explore and summarize effectively. At inference time, the agent requires no external rewards or human instructions. It spontaneously performs native self-evolution to adapt to unknown environments using its internal parameters. When applied to Qwen3-30B and Seed-OSS-36B, this shift to native evolution yields a 20% performance increase on WebVoyager and WebWalker. Most strikingly, the generated world knowledge even enables a compact 14B Qwen3 model to outperform the unassisted Gemini-2.5-Flash, establishing a new paradigm for truly evolving agents.