Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning

2025-11-20

AI Summary

Existing LLM agents rely on human-annotated data for reinforcement learning, suffering from poor scalability and inherent limitations of human knowledge; self-evolution frameworks are often confined to single-turn interactions and static model capabilities, failing to support tool-augmented reasoning and complex, progressive curriculum evolution. Method: We propose a dual-agent symbiotic competition framework, where a single large language model is instantiated as both a curriculum agent and an execution agent. Through multi-round collaborative evolution, the framework autonomously generates hierarchical task curricula and integrates external tools—including mathematical computation and code execution—to enable dynamic reasoning and capability leaps. Evolution is fully autonomous, driven solely by intrinsic rewards without any human-provided data. Contribution/Results: Evaluated on Qwen3-8B-Base, our framework achieves +18% improvement in mathematical reasoning accuracy and +24% in general reasoning performance, significantly advancing autonomous learning and complex problem-solving capabilities of LLM agents.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Evolutionary Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large Language Model (LLM) Agents, often trained with Reinforcement Learning (RL), are constrained by a dependency on human-curated data, limiting scalability and tethering AI to human knowledge. Existing self-evolution frameworks offer an alternative but are typically restricted by the model's inherent capabilities and single-round interactions, hindering the development of complex curricula involving tool use or dynamic reasoning. We introduce Agent0, a fully autonomous framework that evolves high-performing agents without external data through multi-step co-evolution and seamless tool integration. Agent0 establishes a symbiotic competition between two agents initialized from the same base LLM: a curriculum agent that proposes increasingly challenging frontier tasks, and an executor agent that learns to solve them. We integrate external tools to enhance the executor's problem-solving capacity; this improvement, in turn, pressures the curriculum agent to construct more complex, tool-aware tasks. Through this iterative process, Agent0 establishes a self-reinforcing cycle that continuously produces high-quality curricula. Empirically, Agent0 substantially boosts reasoning capabilities, improving the Qwen3-8B-Base model by 18% on mathematical reasoning and 24% on general reasoning benchmarks. Code is available at https://github.com/aiming-lab/Agent0.

Problem

• Developing autonomous agents without human-curated data dependency
• Overcoming limitations of single-round interactions in self-evolution frameworks
• Enhancing reasoning capabilities through tool-integrated multi-step co-evolution

Innovation

• Self-evolving agents via multi-step co-evolution
• Seamless tool integration for enhanced problem-solving
• Symbiotic competition between curriculum and executor agents