Research map / Reinforcement learning / Offline RL
RvS: What is Essential for Offline RL via Supervised Learning?
Scott Emmons, Benjamin Eysenbach, Ilya Kostrikov, Sergey Levine
This paper tried a common change: Add dropout
Add dropout regularization · Kitchen complete · evaluation return Improved
What 795 accepted papers report for this change
- 549 improved
- 87 no clear effect
- 12 mixed
- 147 hurt
- Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference ICML 2026 · apply layer dropout to whole transformer blocks instead of attention and FFN sub-blocks separately · 20 TPP val loss, 271M · val loss: 2.377 -> 2.367Improved
- Quantifying the Uncertainty of Foundation Models with Singular Value Ensembles ICML 2026 · use MC Dropout · vision and SST-2 · accuracy: 86.3 -> 76.6 on Flowers102Hurt
- Motion Planning in Compressed Representation Spaces ICML 2026 · add nested dropout during training · WOMD guided generation · goal FDE: 4.145 -> 1.104Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 |
|---|---|---|---|
| Condition on goal states instead of reward values | RvS on AntMaze v2 · AntMaze v2 average | normalized score: 25.6 -> 53.5 | Improved |
| Condition on goal states instead of reward values | RvS on Kitchen v0 · Kitchen v0 average | normalized score: 1.0 -> 54.0 | Improved |
| Use reward conditioning rather than goal conditioning | RvS on Gym v2 · Gym v2 | normalized score | Improved |
| Use a simple MLP with supervised likelihood training instead of a Transformer sequence model | RvS-R · Gym v2 average | normalized score: 54.7 vs 57.0 | No clear effect |
| Use RvS rather than the strongest value-based baseline | RvS-G · AntMaze v2 average | normalized score: 50.6 -> 53.5 | Improved |
| Use RvS rather than the strongest value-based baseline | RvS-G · Kitchen v0 average | normalized score: 48.2 -> 54.0 | Improved |
| Use reward-conditioned imitation rather than CQL | RvS-R · Gym random-v2 average | normalized score: 10.1 -> 3.8 | Hurt |
| Add dropout regularization (ablation) | RvS · Kitchen complete | evaluation return | Improved |
| Add dropout regularization (ablation) | RvS · Hopper medium-expert and AntMaze medium-play | evaluation return | Hurt |
| Add dropout regularization (ablation) | RvS · Pusher | evaluation return | No clear effect |
| Use a categorical distribution over discretized actions instead of a unimodal Gaussian (ablation) | RvS-G · GCSL | goal success score | Improved |
| Increase network capacity (ablation) | RvS · GCSL and D4RL tasks | evaluation return | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Decision Transformer — Decision Transformer: Reinforcement Learning via Sequence Modeling (NeurIPS 2021) · better on 0 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
- Instabilities of Offline RL with Pre-Trained Neural Representation (ICML 2021 · similarity 0.89)
- Representation Matters: Offline Pretraining for Sequential Decision Making (ICML 2021 · similarity 0.88)
- Are Expressive Models Truly Necessary for Offline RL? (AAAI 2025 · similarity 0.88)
- Tackling Data Corruption in Offline Reinforcement Learning via Sequence Modeling (ICLR 2025 · similarity 0.88)
- A Tractable Inference Perspective of Offline RL (NeurIPS 2024 · similarity 0.87)
- Is Value Learning Really the Main Bottleneck in Offline RL? (NeurIPS 2024 · similarity 0.86)
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.
