AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

MobILE: Model-Based Imitation Learning From Observation Alone

Rahul Kidambi, Jonathan Daniel Chang, Wen Sun

NeurIPS 2021 · Reinforcement learning › Imitation learning · approach: Behavior cloning theory and models · 3 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
Remove optimism from the optimistic MDP construction (ablation)MobILE · MuJoCo suitesample efficiencyHurt
Increase expert trajectories from 5 to 10 (ablation)MobILE · Cartpole-v1expert normalized score: 1.07\pm0.15 -> 1.14\pm0Improved
Increase expert trajectories from 10 to 20 (ablation)MobILE · Reacher-v2expert normalized score: 1.01\pm0.05 -> 0.997\pm0.055No clear effect
Increase expert trajectories from 10 to 40 (ablation)MobILE · Swimmer-v2expert normalized score: 1.54\pm1.1 -> 1.25\pm0.15Improved
Increase expert trajectories from 10 to 40 (ablation)MobILE · Hopper-v2expert normalized score: 1.11\pm0.064 -> 1.16\pm0.03Improved
Increase expert trajectories from 10 to 50 (ablation)MobILE · Walker2d-v2expert normalized score: 0.975\pm0.12 -> 0.94\pm0.038No clear effect
MobILE without expert actions versus BC and GAIL with actionsMobILE · MuJoCo suiteexpert normalized scoreNo clear effect
MobILE versus ILFO baselines BC-O, ILPO, GAIFOMobILE · MuJoCo suiteexpert normalized scoreImproved
MobILE versus BC under stochastic dynamicsMobILE · Stochastic Cartpoletask successImproved

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