Research map / Reinforcement learning / Imitation learning
Imitation Learning by Estimating Expertise of Demonstrators
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Model each demonstrator's expertise when learning from mixed-quality demonstrations | Imitation learning on Robomimic Square · Robomimic Square, Total | final reward: 61.6±3.3 -> 66.4±3.4 | Helped |
| Model demonstrator expertise in the multi-skill setting | MiniGrid multi-skill · β=0.01 | mean episodic reward: 0.44±0.04 -> 0.70±0.06 | Helped |
| Model demonstrator expertise in the multi-skill setting | MiniGrid multi-skill · β=0.50 | mean episodic reward: 0.90±0.01 -> 0.86±0.04 | Hurt |
| Account for demonstrator expertise | Imitation learning on MiniGrid Obstacles · β-1, β-5, β-10, and β-unif | mean episodic reward: 0.18 -> 0.91; 0.66 -> 0.94; 0.63 -> 0.94; 0.80 -> 0.90 | Helped |
| Account for demonstrator expertise | Chess endgame imitation learning · Chess endgames | average pawn loss per move: -4.00 ± 0.11 -> -3.27 ± 0.09 | Helped |
| Increase the number of ILEED training restarts | ILEED · MiniGrid Obstacles | probability of exceeding mean demonstrator performance: 0.37 -> 0.77 -> 0.84 | Helped |
| Remove state embeddings (use state-independent expertise) (ablation) | ILEED · MiniGrid multi-skill, β=.01 | mean episodic reward: 0.70 -> 0.19 | Hurt |
| Remove demonstrator identities (ablation) | ILEED · MiniGrid multi-skill, β=.01 | mean episodic reward: 0.70 -> 0.10 | Hurt |
| Remove state embeddings (use state-independent expertise) (ablation) | ILEED · MiniGrid multi-skill, β=.20 | mean episodic reward: 0.86 -> 0.74 | Hurt |
| Remove demonstrator identities (ablation) | ILEED · MiniGrid multi-skill, β=.20 | mean episodic reward: 0.86 -> 0.64 | Hurt |
| Add state-dependent expertise (ablation) | ILEED on the first three experiments · MiniGrid and Robomimic experiments | performance | No clear effect |
| Add the DeepMDP auxiliary loss (ablation) | ILEED on MiniGrid · MiniGrid experiments | performance | No clear effect |
Closest papers by meaning
- Imitation Learning by Reinforcement Learning (ICLR 2022 · similarity 0.92)
- Imitation Learning from Vague Feedback (NeurIPS 2023 · similarity 0.91)
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning (NeurIPS 2025 · similarity 0.90)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.90)
- Markov Balance Satisfaction Improves Performance in Strictly Batch Offline Imitation Learning (AAAI 2025 · similarity 0.90)
- Efficient Imitation under Misspecification (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.