Research map / Reinforcement learning / Offline RL
False Correlation Reduction for Offline Reinforcement Learning
Zhi‐Hong Deng, Zuyue Fu, Lingxiao Wang, Zhuoran Yang, Chenjia Bai, Tianyi Zhou, Zhaoran Wang, Jing Bo Jiang
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 annealing behavior cloning regularizer (ablation) | SCORE · hopper-medium | uncertainty stability / return | Improved |
| add BC regularizer on diverse data (ablation) | SCORE · hopper-medium-replay | uncertainty stability | Improved |
| use BC regularizer instead of VAE-fitted behavior policy (Fix or MMD) (ablation) | SCORE · D4RL medium-replay and medium | average return | Improved |
| add uncertainty (pessimism) penalty (ablation) | SCORE · hopper-medium-replay and walker2d-medium-replay | average return | Improved |
| use bootstrapped ensemble instead of MC dropout (ablation) | SCORE · D4RL medium-replay | average return | Improved |
| use ensemble penalty instead of min-Q pessimism (ablation) | SCORE · D4RL medium-replay | average return | Improved |
| penalty on u(s,a), u(s',a') or both (ablation) | SCORE · D4RL hopper/walker2d | average return | No clear effect |
| increase penalty coefficient beta to 2.0 (ablation) | SCORE · hopper/walker2d medium-replay | average return: beta 0.1/0.2/0.5 similar; 2.0 poor | Hurt |
| beta 0.1, 0.2 or 0.5 (ablation) | SCORE · D4RL | average return | No clear effect |
| constant BC weight (gamma_bc 1.0) instead of decay (ablation) | SCORE · medium-replay | average return: gamma_bc 0.96/0.98 vs 1.0 | Hurt |
| gamma_bc 0.96 on medium dataset (ablation) | SCORE · medium | average return: 0.98 best; 0.96 collapses late | Hurt |
| increase ensemble size from M=3 to M=5 (ablation) | SCORE · D4RL | performance stability: M in {3,5,7,9} | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MOPO — MOPO: Model-based Offline Policy Optimization (NeurIPS 2020) · better on 1 of 1 shared result
- MOReL — MOReL: Model-Based Offline Reinforcement Learning (NeurIPS 2020) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Delphic Offline Reinforcement Learning under Nonidentifiable Hidden Confounding (ICLR 2024 · similarity 0.90)
- Selective Uncertainty Propagation in Offline RL (AAAI 2025 · similarity 0.90)
- Dynamic Uncertainty Estimation for Offline Reinforcement Learning (AAAI 2025 · similarity 0.90)
- Fewer May Be Better: Enhancing Offline Reinforcement Learning with Reduced Dataset (ICLR 2025 · similarity 0.90)
- Information-Directed Offline-to-Online Reinforcement Learning (NeurIPS 2026 · similarity 0.89)
- A Unified Principle of Pessimism for Offline Reinforcement Learning under Model Mismatch (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.
