method · first seen 20 Oct 2026
advantage-weighted regression
Also: AWR, advantage-weighted behavior cloning
Behavior cloning variant that weights demonstrated actions by their estimated advantage to favor higher-value behaviors.
Mentions per week
PapersCodeDiscussion
Earliest sources
- AWLS: Advantage-Weighted Latent Steering for One-Step Generative Robotic Manipulation PoliciesPapers · 20 Oct 2026 · Jiahong Tang, Jiaju He, Zhiyuan Yang, Ruyang Han, Haiyue Zhu
- Safe Offline Reinforcement Learning using Behavior Regularisation and Latent Feasibility GuidancePapers · 7 Nov 2026 · Mahesh Keswani, Parth Bhardwaj, Ankur Kumar, Raunak P. Bhattacharyya
- Off-Policy RL with Online Inference Tuning: My Solution to the CoRL 2026 RoboEval ChallengePapers · 8 Nov 2026 · Ilia Larchenko
Latest sources
- Off-Policy RL with Online Inference Tuning: My Solution to the CoRL 2026 RoboEval ChallengePapers · 8 Nov 2026 · Ilia Larchenko
- Safe Offline Reinforcement Learning using Behavior Regularisation and Latent Feasibility GuidancePapers · 7 Nov 2026 · Mahesh Keswani, Parth Bhardwaj, Ankur Kumar, Raunak P. Bhattacharyya
- AWLS: Advantage-Weighted Latent Steering for One-Step Generative Robotic Manipulation PoliciesPapers · 20 Oct 2026 · Jiahong Tang, Jiaju He, Zhiyuan Yang, Ruyang Han, Haiyue Zhu