Omar Khattab

GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

Reasoning-Intensive Regression

Multi-module GRPO: Composing Policy Gradients and Prompt Optimization for Language Model Programs

Cross-Reflect: Empowering Multi-Modal Agents with Joint Reasoning Across Trajectories

Grounding by Trying: LLMs with Reinforcement Learning-Enhanced Retrieval