Baselines#

We provide a number of different baselines spanning different categories of learning from demonstrations research: Behavior Cloning / Supervised Learning, Offline Reinforcement Learning, and Online Learning from Demonstrations. This page is still a WIP as we finish running experiments and establish clear baselines and benchmarking setups.

Behavior Cloning (BC) Baselines

BC Baselines are characterized by supervised learning focused algorithms for learning from demonstrations, without any online interaction with the environment.

Baseline

Code

Results

Paper

Standard Behavior Cloning (BC)

WIP

WIP

N/A

Diffusion Policy (DP)

Link

WIP

Link

Action Chunking Transformer (ACT)

Link

WIP

Link

Online Learning from Demonstrations Baselines

Online learning from demonstrations baselines are characterized by learning from demonstrations while also leveraging online environment interactions.

Baseline

Code

Results

Paper

Reverse Forward Curriculum Learning (RFCL)*

Link

WIP

Link

Reinforcement Learning from Prior Data (RLPD)

Link

WIP

Link

SAC + Demos (SAC+Demos)

WIP

N/A

* - This indicates the baseline uses environment state reset which is typically a simulation only feature