Research map / Reinforcement learning / Imitation learning
Adversarial Dual On-Policy Distillation from Expressive Teacher
Zhenglin Wan, Jingxuan Wu, Xingrui Yu, Chubin Zhang, Mingcong Lei, Bo An, Ivor Tsang, Yang You
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 FA-OPD online distillation to flow/diffusion behavioral cloning (expert return 1965.58) | Flow-Matching Policy · Walker2d sub-optimal expert ID 2 | average return: 1342.47 -> 2093.14 | Improved |
| add FA-OPD online distillation with near-novice expert (return 265.36) | Diffusion Policy · Walker2d sub-optimal expert ID 1 | average return: 247.24 -> 99.05 | Hurt |
| add FA-OPD online distillation with strong expert (return 5357.37) | Diffusion Policy · Walker2d sub-optimal expert ID 5 | average return: 2091.17 -> 4287.24 | Improved |
| add online interaction to DP/FP under 1.5x noise | Flow-Matching Policy · Hand-rotate noise generalization | success rate: 0.91 to nearly 0.82 (DP/FP) vs above 0.95 (FA-OPD) | Improved |
| student distilled from FM teacher instead of updating FM policy online with PPO | FM-PPO · Maze2d | success rate: peak near 0.33 then declines | Improved |
| add action-distillation channel to reward-only distillation (beta>0) (ablation) | FA-OPD · ablation over beta | performance | Improved |
Closest papers by meaning
- Mimicking Better by Matching the Approximate Action Distribution (ICML 2024 · similarity 0.87)
- Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback (NeurIPS 2026 · similarity 0.87)
- Off-Dynamics Reinforcement Learning via Domain Adaptation and Reward Augmented Imitation (NeurIPS 2024 · similarity 0.87)
- Non-Adversarial Inverse Reinforcement Learning via Successor Feature Matching (ICLR 2025 · similarity 0.87)
- The Many Faces of On-Policy Distillation: Pitfalls, Mechanisms, and Fixes (NeurIPS 2026 · similarity 0.87)
- Implicit Drifting Policy: One-Step Action Generation via Conditional Expert Geometry (NeurIPS 2026 · 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.
