Research map / Reinforcement learning / Offline RL
Decision Transformer: Reinforcement Learning via Sequence Modeling
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 return-to-go conditioning (ablation) | behavior cloning transformer · Atari Breakout | gamer-normalized score: 138.9 -> 267.5 | Helped |
| Decision Transformer vs CQL | offline RL · D4RL average | normalized score: 63.9 -> 74.7 | Helped |
| Decision Transformer vs CQL | offline RL · Atari Qbert | gamer-normalized score: 104.2 -> 15.1 | Hurt |
| Decision Transformer vs best percentile BC | offline RL · D4RL | normalized score: 56.7 -> 56.1 | No clear effect |
| Decision Transformer vs percentile BC in low-data regime | offline RL · Atari Breakout | gamer-normalized score: 108.2 -> 267.5 | Helped |
| sequence modeling vs TD learning for long-horizon credit assignment | offline RL · Key-to-Door 1K | success rate: 13.1% -> 71.8% | Helped |
| replace dense rewards with delayed (sparse) returns | Decision Transformer · D4RL Hopper Medium-Replay | normalized score: 82.7 -> 78.5 | No clear effect |
| replace dense rewards with delayed (sparse) returns | CQL · D4RL Hopper Medium-Replay | normalized score: 48.6 -> 2.0 | Hurt |
| use MLP with K=1 context instead of transformer (ablation) | Decision Transformer · offline RL | normalized score | Hurt |
Closest papers by meaning
- Rethinking Decision Transformer via Hierarchical Reinforcement Learning (ICML 2024 · similarity 0.93)
- Decision ConvFormer: Local Filtering in MetaFormer is Sufficient for Decision Making (ICLR 2024 · similarity 0.91)
- Decision Mamba: Reinforcement Learning via Hybrid Selective Sequence Modeling (NeurIPS 2024 · similarity 0.91)
- In-Context Decision Transformer: Reinforcement Learning via Hierarchical Chain-of-Thought (ICML 2024 · similarity 0.90)
- In-Context Reinforcement Learning From Suboptimal Historical Data (ICML 2025 · similarity 0.90)
- Transformers as Decision Makers: Provable In-Context Reinforcement Learning via Supervised Pretraining (ICLR 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.