Policy Gradient Experiments

Collection of policy gradient and reinforcement learning experiments for various domains.

Experiments

Branch	Description	Domain
MBPP	Code generation with GRPO + vLLM + SandboxFusion	MBPP coding tasks
gsm8k	Mathematical reasoning with GRPO	GSM8K dataset
math_posttrain	Post-training experiments	MATH dataset (Hendrycks)

Each experiment lives in its own branch with complete code, documentation, and results. The main branch serves as an index to all experiments.