BiGym

The BiGym dataset is a benchmark that contains a variety of robotic manipulation tasks used to evaluate reinforcement learning methods, particularly in the context of long-horizon tasks with sparse rewards.

Papers using BiGym (3)

Actor-Critic for Continuous Action Chunks: A Reinforcement Learning Framework for Long-Horizon Robotic Manipulation with Sparse Reward2025Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning2024BiGym: A Demo-Driven Mobile Bi-Manual Manipulation Benchmark2024