Research map / Reinforcement learning / Visual and representation RL
In-context Reinforcement Learning with Algorithm Distillation
Michael Laskin, Luyu Wang, Junhyuk Oh, Emilio Parisotto, Stephen Spencer, Richie Steigerwald, DJ Strouse, Steven Stenberg Hansen, Angelos Filos, Ethan Brooks, maxime gazeau, Himanshu Sahni and 2 more
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 across-episode causal transformer distillation of source RL learning histories (AD) vs expert-only data (ED) | AD · Dark Room, Dark Room Hard, Dark Key-to-Door, Watermaze | return | Improved |
| Train on expert trajectories only (Expert Distillation) instead of learning histories (ablation) | ED · Dark Room, Dark Room Hard, Watermaze | return | Hurt |
| Distill from multi-stream learning histories into single-stream AD (data-efficiency vs source) | AD · Dark environments and Watermaze | sample efficiency vs source algorithm | Improved |
| Increase across-episode context length to 2-4 episodes (ablation) | AD · Dark Room | return | Improved |
| Replace transformer with LSTM in AD (ablation) | AD · Dark Key-to-Door | return | Hurt |
| Add label smoothing (alpha 0.2) to AD (ablation) | AD · Dark Room (Hard) | in-context learning performance | Improved |
| Pre-fill context with demonstration policy data | AD · Dark Room | return | Improved |
| Pre-fill context with demonstration policy data | ED · Dark Room | return | No clear effect |
| Use RL2 (online meta-RL) as upper bound reference | RL2 · Dark Room, Watermaze | return | No clear effect |
Closest papers by meaning
- Distilling Reinforcement Learning Algorithms for In-Context Model-Based Planning (ICLR 2025 · similarity 0.90)
- Vintix: Action Model via In-Context Reinforcement Learning (ICML 2025 · similarity 0.89)
- In-Context Decision Transformer: Reinforcement Learning via Hierarchical Chain-of-Thought (ICML 2024 · similarity 0.88)
- XLand-100B: A Large-Scale Multi-Task Dataset for In-Context Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Hierarchical Reinforcement Learning with Targeted Causal Interventions (ICML 2025 · similarity 0.88)
- Towards Provable Emergence of In-Context Reinforcement Learning (NeurIPS 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.
