Research map / Reinforcement learning / Imitation learning
Learning from Interventions Using Hierarchical Policies for Safe Learning
Jing Bi, Vikas Dhiman, Tianyou Xiao, Chenliang Xu
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 |
|---|---|---|---|
| Use hierarchical sub-goal policies instead of a direct Branched policy | LfI in CARLA · CARLA urban driving | time and distance without intervention | Improved |
| Add Backtracking to interpolate actions around an expert intervention | LfI in CARLA · CARLA urban driving | time and distance without intervention | Improved |
| Combine hierarchical sub-goal policies with Backtracking | LfI in CARLA · CARLA urban driving | time and distance without intervention | Improved |
| Use sub-goals with LbB rather than Branched+Demo | LfI in CARLA · CARLA urban driving | training data samples per iteration | Improved |
| Increase the future sub-goal horizon from a small value toward k=5 (ablation) | Sub-goal+LbB · CARLA urban driving | time and distance without intervention | Improved |
| Increase the future sub-goal horizon beyond k=5 (ablation) | Sub-goal+LbB · CARLA urban driving | time and distance without intervention | Hurt |
| Use Branched rather than Sub-goal policy representation (ablation) | LfI in CARLA · CARLA urban driving | time and distance without intervention | Hurt |
| Use Demo rather than LbB data collection (ablation) | LfI in CARLA · CARLA urban driving | training data samples per iteration | Hurt |
Closest papers by meaning
- Learning When to Stop: Selective Imitation Learning Under Arbitrary Dynamics Shift (NeurIPS 2026 · similarity 0.87)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.86)
- Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback (NeurIPS 2026 · similarity 0.86)
- When Does Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning (NeurIPS 2026 · similarity 0.86)
- Demonstration-Regularized RL (ICLR 2024 · similarity 0.86)
- How to Leverage Diverse Demonstrations in Offline Imitation Learning (ICML 2024 · similarity 0.85)
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.
