Research map / Reinforcement learning / Imitation learning
CEIP: Combining Explicit and Implicit Priors for Reinforcement Learning with Demonstrations
Kai Yan, Alex Schwing, Yu-Xiong 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 |
|---|---|---|---|
| Combine explicit retrieval with the learned implicit flow prior (ablation) | CEIP · Kitchen-SKiLD and Kitchen-FIST | reward | Improved |
| Add explicit retrieval conditioning (ablation) | PARROT · Kitchen and Office | reward | Improved |
| Add push-forward retrieval to explicit-prior conditioning (ablation) | CEIP and PARROT · Office | reward | Improved |
| Add task-specific flow for subtasks missing from task-agnostic demonstrations (ablation) | CEIP · Kitchen-FIST | reward | Improved |
| Add task-specific flow (ablation) | CEIP · Office | reward | No clear effect |
| Add explicit prior and push-forward retrieval (ablation) | CEIP · Fetchreach | reward growth | Hurt |
| Add task-specific flow (ablation) | CEIP · Fetchreach | reward | No clear effect |
| Increase the number of combined task-agnostic flows within a reasonable range (ablation) | CEIP · Fetchreach | reward | Improved |
| Use more task-relevant task-agnostic demonstrations (ablation) | CEIP and PARROT · Demonstration-guided RL | reward | Improved |
| Use ground-truth task labels for clustering (ablation) | CEIP · Demonstration-guided RL | reward | Mixed |
| Add explicit prior to behavior cloning (ablation) | Behavior cloning · Demonstration-guided RL | reward | Improved |
| Add explicit prior to FIST-style retrieval (ablation) | FIST · Fetchreach | reward | Hurt |
Closest papers by meaning
- APC-RL: Exceeding data-driven behavior priors with adaptive policy composition (ICLR 2026 · similarity 0.87)
- Learning a Prior over Intent via Meta-Inverse Reinforcement Learning (ICML 2019 · similarity 0.86)
- PsiPhi-Learning: Reinforcement Learning with Demonstrations using Successor Features and Inverse Temporal Difference Learning (ICML 2021 · similarity 0.86)
- Uncertainty-Based Smooth Policy Regularisation for Reinforcement Learning with Few Demonstrations (NeurIPS 2025 · similarity 0.86)
- Sequential Decision Making with Expert Demonstrations under Unobserved Heterogeneity (NeurIPS 2024 · similarity 0.84)
- XLand-100B: A Large-Scale Multi-Task Dataset for In-Context Reinforcement Learning (ICLR 2025 · similarity 0.84)
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.
