D4RL

D4RL is a benchmark that contains a variety of offline reinforcement learning tasks used to evaluate the performance of algorithms in real-world robot learning scenarios.

Papers using D4RL (6)

Neuro-Inspired Inverse Learning for Planning and Control2026Off-policy Actor-critic With Sigmoid-bounded Entropy For Real-world Robot Learning2026Opinion: Towards Unified Expressive Policy Optimization For Robust Robot Learning2025Opinion: Towards Unified Expressive Policy Optimization for Robust Robot Learning2025Off-Policy Actor-Critic with Sigmoid-Bounded Entropy for Real-World Robot Learning2026Two-Steps Diffusion Policy for Robotic Manipulation via Genetic Denoising2025