Research map / Reinforcement learning / Offline RL
Behavior Regularization with Flow Latent Policy for Offline Reinforcement Learning
Yulong Xia, Fuchun Sun
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add best-of-N sampling (N=4) (ablation) | FLP · OGBench humanoid-medium, cube-double | success rate (%) | Improved |
| set KL coefficient alpha to 0.1 or 0.3 instead of moderate values (ablation) | FLP · OGBench representative tasks | success rate (%) | Hurt |
| increase best-of-N samples from 4 to 8 or 16 (ablation) | FLP · OGBench representative tasks | success rate (%) | No clear effect |
| latent-space Gaussian policy on frozen flow instead of one-step distilled flow policy | FQL · OGBench cube-double (offline) | success rate (%): 29 -> 69 | Improved |
| latent-space Gaussian policy on frozen flow instead of one-step distilled flow policy | FQL · OGBench humanoid-m (offline) | success rate (%): 58 -> 72 | Improved |
| latent-space Gaussian policy on frozen flow instead of one-step distilled flow policy | FQL · OGBench antmaze-large (offline) | success rate (%): 79 -> 75 | Hurt |
| latent-space Gaussian policy on frozen flow instead of one-step distilled flow policy | FQL · OGBench cube-single (offline) | success rate (%): 96 -> 93 | No clear effect |
| online fine-tuning with frozen flow, FLP vs FQL | FLP · OGBench puzzle-4x4 (offline-to-online) | success rate (%): 11 -> 99 | Improved |
Closest papers by meaning
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.91)
- Behavior-Regularized Diffusion Policy Optimization for Offline Reinforcement Learning (ICML 2025 · similarity 0.91)
- Constrained Latent Action Policies for Model-Based Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
- Flow-Based Policy for Online Reinforcement Learning (NeurIPS 2025 · similarity 0.90)
- Adaptive Advantage-Guided Policy Regularization for Offline Reinforcement Learning (ICML 2024 · similarity 0.90)
- DeFlow: Decoupling Behavior-Prior Modeling and Value Maximization for Offline Policy Extraction (NeurIPS 2026 · similarity 0.90)
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.
