ExoRL

The 'ExoRL' dataset/benchmark contains continuous control tasks used to evaluate off-policy learning methods in a zero-shot reinforcement learning context.

Papers using ExoRL (4)

TD-JEPA: Latent-predictive Representations for Zero-Shot Reinforcement Learning2025 · 9 citesZero-Shot Off-Policy Learning2026Towards Robust Zero-Shot Reinforcement Learning2025Pretraining a Shared Q-Network for Data-Efficient Offline Reinforcement Learning2025