AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Actionable Models: Unsupervised Offline Reinforcement Learning of Robotic Skills

Yevgen Chebotar, Karol Hausman, Yao Lu, Ted Xiao, Dmitry Kalashnikov, Jacob Varley, Alex Irpan, Benjamin Eysenbach, Ryan C Julian, Chelsea Finn, Sergey Levine

ICML 2021 · Reinforcement learning › Offline RL · approach: Diffusion models for offline decision making · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add goal chaining to Actionable Models (ablation)Actionable Models · Real-world container placingsuccess rate: 8% to 66%Improved
pre-train on offline goal-reaching data before task-reward fine-tuningReal-world instance grasping · Grasp boxsuccess rate: 0% -> 27%Improved
pre-train on offline goal-reaching data before task-reward fine-tuningReal-world instance grasping · Grasp bananasuccess rate: 4% -> 20%Improved
pre-train on offline goal-reaching data before task-reward fine-tuningReal-world instance grasping · Grasp milksuccess rate: 1% -> 20%Improved
remove conservative regularization for unseen actions (ablation)offline goal-conditioned Q-learning · Simulated visual goal reachingsuccess rateHurt
replace contrastive unseen-action negatives with random goal negativesoffline goal-conditioned Q-learning · Simulated visual goal reachingsuccess rateHurt
remove goal chaining (ablation)Actionable Models · Simulated fixture placingsuccess rateHurt
use goal-conditioned pre-trainingQT-Opt online learning · Simulated instance graspingtraining speedImproved
add a 10% auxiliary goal-reaching objectiveQT-Opt online learning · Simulated instance graspingtraining speedImproved
use goal-conditioned behavioral cloning instead of Actionable Modelsoffline simulated visual goal reaching · simulated manipulation taskssuccess rateHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy