Malliavin Calculus for Counterfactual Gradient Estimation in Adaptive Inverse Reinforcement Learning

Create a robust implementation of passive Langevin-based adaptive IRL for practitioners. This could help developers reverse-engineer complex behavioral policies from raw observation logs.

Suggested repo: langevin-irl

"Reverse engineer reward functions from passive observation logs using Langevin dynamics."

Estimated effort: 80h