Code for Agentic Policy Optimization via Instruction-Policy Co-Evolution

💡 Overview

We introduce INSPO, a novel Instruction-Policy Co-Evolution framework that integrates instruction optimization as a dynamic component of the reinforcement learning (RL) loop for training large language models.

📑 Citation

If you find INSPO useful for your research and applications, please give us a star. We will release the scripts in the near future.