Research map / Reinforcement learning / Offline RL
Tackling Heavy-Tailed Q-Value Bias in Offline-to-Online Reinforcement Learning with Laplace-Robust Modeling
Ruibo Guo, Lei Liu, Rui Yang, Junjie Shen, Guoping Wu, Jie Wang, Bin Li
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 |
|---|---|---|---|
| use LAROO instead of BOORL | offline-to-online RL · D4RL MuJoCo | average score: +54.8% | Improved |
| use LAROO instead of BOORL | offline-to-online RL · walker2d-random | normalized return: 6.4 -> 71.6 | Improved |
| use LAROO instead of BOORL | offline-to-online RL · D4RL MuJoCo, all tasks | delta_sum(0.1M): 355.4 -> 550.4 | Improved |
| use LAROO instead of BOORL | offline-to-online RL · halfcheetah-random | normalized return: 90.6 -> 87.4 | Hurt |
| use LAROO instead of BOORL | offline-to-online RL · hopper-expert | normalized return: 103.2 -> 112.0 | Improved |
| replace l2 loss with Laplace-based Db(x) loss in TD3, Cal-QL and ENOTO | TD3, Cal-QL, ENOTO · walker2d tasks | normalized return | Improved |
| add ensemble models to LAROO (ablation) | LAROO · walker2d-medium-replay | Q bias distribution and return | Improved |
| add Laplace noise model to LAROO (ablation) | LAROO · walker2d-medium-replay | Q bias distribution and return | Improved |
| compare Laplace noise model against ensemble models (ablation) | LAROO · walker2d-medium-replay | performance | Improved |
| use LAROO with UTD=1 and ensemble size 1 instead of Cal-QL (ablation) | LAROO · D4RL, 13 experiments | normalized return: 10 out of 13 | Improved |
| use Lb(x) loss instead of Huber and Cauchy losses (ablation) | LAROO · D4RL | normalized return | Improved |
| model Q bias with Laplace instead of Gaussian (ablation) | LAROO noise model · walker2d-medium-expert-v2 | fit to empirical Q bias | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Cal-QL — Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning (NeurIPS 2023) · 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
- A Perspective of Q-value Estimation on Offline-to-Online Reinforcement Learning (AAAI 2024 · similarity 0.93)
- Towards Robust Offline Reinforcement Learning under Diverse Data Corruption (ICLR 2024 · similarity 0.90)
- Exclusively Penalized Q-learning for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Q-Distribution guided Q-learning for offline reinforcement learning: Uncertainty penalized Q-value via consistency model (NeurIPS 2024 · similarity 0.89)
- Optimistic Critic Reconstruction and Constrained Fine-Tuning for General Offline-to-Online RL (NeurIPS 2024 · similarity 0.89)
- Robust Policy Expansion for Offline-to-Online RL under Diverse Data Corruption (NeurIPS 2025 · 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.
