Self-improving agents

Sources

• https://lean-lang.org/
• https://github.com/deepseek-ai/DeepSeek-Math-V2
  • An approach to build training data set based on existing LLM
  • Self-verifiable reasononing: examiner, generator and supervisor
    • generator provides solution and assesses it
• https://github.com/aiming-lab/Agent0/tree/main/Agent0
  • Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning