✍️ Blog
-
2026.10.20
Flow Matching Policy Optimization with Mirror Descent and Entropy ConstraintsWhy ODE-based flow matching makes expressive RL policies trainable without backpropagating through the sampler, and with an entropy you can actually compute.