Research map / Reinforcement learning / Offline RL
RAD: Retrieval High-quality Demonstrations to Enhance Decision-making
Lu Guo, Yixiang Shan, Zhengbang Zhu, Qifan Liang, Lichang Song, Ting Long, Weinan Zhang, Yi Chang
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 retrieval of high-return target states, step estimation and planning to DiffuserLite | DiffuserLite · AntMaze (average) | normalized score: 77.6 -> 83.3 | Improved |
| add retrieval-augmented targets to DiffuserLite | DiffuserLite · Gym-MuJoCo locomotion (average) | normalized score: 86.0 -> 87.5 | Improved |
| add retrieval-augmented targets to Diffuser | Diffuser · Maze2D (average) | normalized score: 119.5 -> 134.9 | Improved |
| add retrieval-augmented targets to DiffuserLite | DiffuserLite · FrankaKitchen (average) | normalized score: 74.0 -> 72.1 | Hurt |
| add retrieval-augmented targets to Diffuser | Diffuser · Medium-Expert locomotion | normalized score | No clear effect |
| add retrieval-augmented targets to DiffuserLite under region-removed shift | DiffuserLite · AntMaze-Medium-Play, 0% coverage | normalized score: 0.0 -> 36.7 | Improved |
| add retrieval-augmented targets to DiffuserLite under 20% coverage | DiffuserLite · AntMaze-Large-Play, 20% coverage | normalized score: 0.0 -> 47.8 | Improved |
| add retrieval-augmented targets to DiffuserLite under 0% coverage | DiffuserLite · AntMaze-Large-Play, 0% coverage | normalized score: 0.0 -> 0.0 | No clear effect |
| add retrieval targets to DiffuserLite under shifted initial states | DiffuserLite · Hopper, Random initial states | normalized score: 60.4 -> 90.2 | Improved |
| add retrieval targets vs MOPO under shifted initial states | DL-RAD · HalfCheetah, Random initial states | normalized score: 62.3 -> 39.2 | Hurt |
| remove target selection module (-TS) (ablation) | D-RAD · Medium-Expert/Medium/Medium-Replay MuJoCo | normalized score | Improved |
| replace retrieved targets with random targets (-TSR) (ablation) | D-RAD · MuJoCo locomotion | normalized score | 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.
- MOPO — MOPO: Model-based Offline Policy Optimization (NeurIPS 2020) · 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
- ContraDiff: Planning Towards High Return States via Contrastive Learning (ICLR 2025 · similarity 0.89)
- Enhancing Online Reinforcement Learning with Meta-Learned Objective from Offline Data (AAAI 2025 · similarity 0.87)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.87)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.87)
- Offline Inverse RL: New Solution Concepts and Provably Efficient Algorithms (ICML 2024 · similarity 0.87)
- Offline RL with Observation Histories: Analyzing and Improving Sample Complexity (ICLR 2024 · similarity 0.87)
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.
