AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Fighting Copycat Agents in Behavioral Cloning from Observation Histories

Chuan Wen, Jierui Lin, Trevor Darrell, Dinesh Jayaraman, Yang Gao

NeurIPS 2020 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Imitation and skill learning

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
use observation histories instead of a single observationbehavioral cloning · MuJoCo partially observed taskscumulative rewardMixed
add target-conditioned adversarial prediction of the previous actionbehavioral cloning with observation histories · MuJoCo partially observed taskscumulative reward: 592 -> 1296Improved
add target-conditioned adversarial prediction of the previous actionbehavioral cloning with observation histories · six MuJoCo environmentsaction-prediction MSE given past actions: 0.66±0.04 -> 2.20 ± 0.06 (Ant); 1.07±0.16 -> 2.26 ± 0.12 (Hopper); 0.18±0.01 -> 1.33 ± 0.02 (Humanoid); 0.32±0.05 -> 2.9Improved
add an information bottleneck without an adversary (ablation)behavioral cloning with observation histories · MuJoCo partially observed taskscumulative rewardImproved
add target-conditioning to the adversary (ablation)adversarial behavioral cloning · MuJoCo partially observed taskscumulative rewardImproved
add target-conditioning to the adversary (ablation)adversarial behavioral cloning · Hoppernext-action prediction test MSE: 14.62 ± 1.41 -> 9.48 ± 1.89Improved
add an information bottleneck to TCA (ablation)adversarial behavioral cloning · MuJoCo partially observed taskscumulative reward: Ant: 1970 ± 107 -> 2150 ±34; Hopper: 683 ± 132 -> 1086 ±262; Humanoid: 696 ±48 -> 671 ±61; Reacher: -57 ± 4 -> -54 ±4; WMixed
remove target-conditioning from the adversary (ablation)adversarial behavioral cloning · Hopperprevious-action prediction test MSE: 5.97 ± 1.32Hurt
add a target-conditioned adversary and information bottleneckBC-OH · MuJoCo partially observed taskscumulative reward: 592 -> 1296 (Walker2d)Mixed
use Dropout-BC on historical observationsbehavioral cloning with observation histories · MuJoCo partially observed taskscumulative rewardHurt
increase DAGGER expert queries from 100 to 1kDAGGER · MuJoCo partially observed taskscumulative reward: 2090 -> 2240 (Ant); 978 -> 1120 (Hopper); 688* -> 812* (Humanoid); -52 -> -15 (Reacher); 701 -> 2170 (Walker2d); 1080 ->Improved
use observation histories rather than a single observationbehavioral cloning · MuJoCo partially observed tasksheld-out demonstration test loss: 0.1079 -> 0.0829 (Ant); 0.0077 -> 0.0049 (Hopper); 0.5710 -> 0.5470 (Humanoid); 0.0008 -> 0.0007 (Reacher); 0.0410 -> 0.Improved

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