We introduce ExploreNet🔍 for learnable exploration in diffusion RL. ExploreNet learns an adaptive exploration distribution conditioned on the current state, rewarded by the diversity of the rollouts. ➡️ Faster, more targeted learning in diffusion GRPO.