Research map / Reinforcement learning / Visual and representation RL
Scalable In-Context Q-Learning
Jinmei Liu, Fuhong Liu, Zhenhong Sun, Jianye HAO, Huaxiong Li, Bo Wang, Daoyi Dong, Chunlin Chen, Zhi Wang
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 world-model prompt construction (ablation) | S-ICQL · Ant-Dir (Mixed) | return: 693.87 -> 813.34 | Improved |
| add in-context Q-learning (value functions with AWR) (ablation) | S-ICQL · Ant-Dir (Mixed) | return: 784.07 -> 813.34 | Improved |
| add both world model and Q-learning to DPT (ablation) | DPT · Cheetah-Vel (Mixed) | return: -78.35 -> -35.48 | Improved |
| add world model prompt (ablation) | S-ICQL · Reach (Mixed) | return: 792.09 -> 806.97 | Improved |
| lower AWR temperature lambda from 1.0 to 0.01 (ablation) | S-ICQL · Ant-Dir (Mixed) | return: 486.23 -> 813.34 | Improved |
| lower lambda from 0.01 to 0.001 (ablation) | S-ICQL · Ant-Dir (Mixed) | return: 813.34 -> 767.55 | Hurt |
| increase context steps k from 4 to 6 (ablation) | S-ICQL · Ant-Dir (Mixed) | return: 813.34 -> 648.70 | Hurt |
| Q-learning stitching on suboptimal data | S-ICQL · training tasks, suboptimal data | return vs best dataset return | Improved |
| use S-ICQL instead of DIT/IC-IQL baselines | S-ICQL · Walker-Param (Mixed) | return: 391.38 -> 466.72 | Improved |
Closest papers by meaning
- Vintix: Action Model via In-Context Reinforcement Learning (ICML 2025 · similarity 0.92)
- XLand-100B: A Large-Scale Multi-Task Dataset for In-Context Reinforcement Learning (ICLR 2025 · similarity 0.91)
- Towards Large-Scale In-Context Reinforcement Learning by Meta-Training in Randomized Worlds (NeurIPS 2025 · similarity 0.91)
- Zero-shot Model-based Reinforcement Learning using Large Language Models (ICLR 2025 · similarity 0.90)
- In-Context Decision Transformer: Reinforcement Learning via Hierarchical Chain-of-Thought (ICML 2024 · similarity 0.90)
- Q-SFT: Q-Learning for Language Models via Supervised Fine-Tuning (ICLR 2025 · similarity 0.90)
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.
