Research map / Reinforcement learning / Offline RL
Keep Doing What Worked: Behavior Modelling Priors for Offline Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Replace the raw behavior-model prior with the advantage-weighted behavior-model prior (ablation) | MPO offline RL · Control Suite: Cheetah, Hopper, Quadruped, Walker | episode return: 891.8 -> 907.4; 453.8 -> 539.3; 710.8 -> 786.6; 817.7 -> 843.0 | Mixed |
| Replace the raw behavior-model prior with the advantage-weighted behavior-model prior (ablation) | MPO offline RL · Simulated robot tasks | return: 178.3 -> 182.3; 114.6 -> 123.7; 144.2 -> 147.1; 124.4 -> 137.9; 111.3 -> 126.3; 86.9 -> 109.9; 53.5 -> 77.9 | Helped |
| Add policy improvement to the ABM prior (rather than using the prior alone) (ablation) | ABM offline RL · Control Suite | episode return: 888.7 -> 907.4; 454.8 -> 539.3; 733.4 -> 786.6; 835.9 -> 843.0 | Mixed |
| Use ABM with epsilon = 0 and trajectory length 2, omitting policy improvement (ablation) | ABM offline RL · Control Suite | episode return: 884.3; 451.8; 733.6; 838.1 | Mixed |
| Use ABM instead of the raw behavior-model prior (ablation) | MPO offline RL · Control Suite and simulated robot tasks | return | Helped |
Closest papers by meaning
- Continuous Doubly Constrained Batch Reinforcement Learning (NeurIPS 2021 · similarity 0.90)
- Behavior Estimation from Multi-Source Data for Offline Reinforcement Learning (AAAI 2023 · similarity 0.90)
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
- Video-Enhanced Offline Reinforcement Learning: A Model-Based Approach (ICML 2025 · similarity 0.89)
- A Clean Slate for Offline Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Offline Behavior Distillation (NeurIPS 2024 · similarity 0.89)
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.