Research map / Reinforcement learning / Offline RL
Data-Driven Offline Decision-Making via Invariant Representation Learning
Han Qi, Yi Su, Aviral Kumar, Sergey Levine
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 invariant representation regularization (IOM) instead of the compared prior methods | Offline model-based optimization · Design-Bench tasks (aggregate) | normalized score | Improved |
| Add a conservatism term to IOM (IOM-C) (ablation) | IOM · Design-Bench tasks (aggregate) | normalized score | No clear effect |
| Use chi-squared discrepancy rather than MMD (ablation) | IOM · Design-Bench tasks | normalized score | Improved |
| Tune IOM with validation prediction error and an invariance filter | IOM · Design-Bench tasks (aggregate) | normalized score relative to oracle tuning | Improved |
| Tune IOM using validation prediction error only (ablation) | IOM · Design-Bench tasks (aggregate) | normalized score relative to oracle tuning | Hurt |
| Tune IOM using the invariance regularizer only (ablation) | IOM · Design-Bench tasks (aggregate) | normalized score relative to oracle tuning | Hurt |
| Use uniform checkpoint selection instead of validation-error checkpoint selection (ablation) | IOM offline tuning · Design-Bench tasks | objective value | Hurt |
| Select checkpoints using validation prediction error | IOM · Design-Bench tasks | objective value | Improved |
| Vary the quantile used to filter models by invariance (ablation) | IOM offline tuning · Design-Bench tasks (aggregate) | normalized score relative to oracle tuning | No clear effect |
| Use offline validation-based tuning rather than oracle online tuning (ablation) | COMs and gradient-ascent objective models · Design-Bench tasks (aggregate) | normalized score relative to oracle tuning | Hurt |
Closest papers by meaning
- Representation Matters: Offline Pretraining for Sequential Decision Making (ICML 2021 · similarity 0.91)
- Reining Generalization in Offline Reinforcement Learning via Representation Distinction (NeurIPS 2023 · similarity 0.90)
- Improving Generalization in Offline Reinforcement Learning via Latent Distribution Representation Learning (AAAI 2025 · similarity 0.90)
- Tackling Data Corruption in Offline Reinforcement Learning via Sequence Modeling (ICLR 2025 · similarity 0.89)
- Beyond OOD State Actions: Supported Cross-Domain Offline Reinforcement Learning (AAAI 2024 · similarity 0.89)
- Fewer May Be Better: Enhancing Offline Reinforcement Learning with Reduced Dataset (ICLR 2025 · similarity 0.88)
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.
