AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Learning a Prior over Intent via Meta-Inverse Reinforcement Learning

Kelvin Xu, Ellis Ratner, Anca Dragan, Sergey Levine, Chelsea Finn

ICML 2019 · Reinforcement learning › Imitation learning · approach: Theory of inverse RL · 8 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
Meta-train a reward initialization for fast adaptation rather than train from scratchMaxEnt IRL on SUNCG · SUNCG Test (5 demonstrations)success rate (%): 65.4 -> 77.3Improved
Meta-train a reward initialization for fast adaptation rather than train from scratchMaxEnt IRL on SUNCG · SUNCG Unseen Houses (5 demonstrations)success rate (%): 76.5 -> 82.6Improved
Optimize the initialization for adaptation instead of following the average task gradientMaxEnt IRL on SUNCG · SUNCG Test (5 demonstrations)success rate (%): 60.8 -> 77.3Improved
Optimize the initialization for adaptation instead of following the average task gradientMaxEnt IRL on SUNCG · SUNCG Unseen Houses (5 demonstrations)success rate (%): 73.3 -> 82.6Improved
Adapt the meta-learned reward initialization using demonstrations (ablation)MandRIL on SUNCG · SUNCG Test (5 demonstrations)success rate (%): 20.7 -> 77.3Improved
Adapt the meta-learned reward initialization using demonstrations (ablation)MandRIL on SUNCG · SUNCG Unseen Houses (5 demonstrations)success rate (%): 25.3 -> 82.6Improved
Use average-gradient supervised pretraining instead of random initializationMaxEnt IRL on SUNCG · SUNCG Test (5 demonstrations)success rate (%): 65.4 -> 60.8Hurt
Use average-gradient supervised pretraining instead of random initializationMaxEnt IRL on SUNCG · SUNCG Unseen Houses (5 demonstrations)success rate (%): 76.5 -> 73.3Hurt
Use an LSTM-based recurrent meta-learner instead of MandRILSpriteWorld reward learning · Held-out tasks and held-out tasks with novel spritesvalue differenceHurt
Increase the number of demonstrations for learning from scratchMaxEnt IRL on SpriteWorld · Held-out tasksvalue difference: 20 or more demonstrationsMixed
Apply bias transformation (ablation)SpriteWorld LSTM meta-learner · SpriteWorldperformanceNo clear effect
Condition the LSTM initial hidden state on image features (ablation)SpriteWorld LSTM meta-learner · SpriteWorldperformanceNo clear effect

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