AgentFactory shifts agent evolution from unreliable textual 'reflections' to a library of verifiable, executable Python subagents.
March 19, 2026
Original Paper
AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse
arXiv · 2603.18000
The Takeaway
Current agents rely on long-context prompts or textual memories that often fail in complex scenarios. This framework enables agents to accumulate and refine a permanent library of robust Python code solutions, allowing for reliable task re-execution and portability across systems.
From the abstract
Building LLM-based agents has become increasingly important. Recent works on LLM-based agent self-evolution primarily record successful experiences as textual prompts or reflections, which cannot reliably guarantee efficient task re-execution in complex scenarios. We propose AgentFactory, a new self-evolution paradigm that preserves successful task solutions as executable subagent code rather than textual experience. Crucially, these subagents are continuously refined based on execution feedback