Research map / Reinforcement learning / Imitation learning
Imitation Learning from Imperfection: Theoretical Justifications and Algorithms
Ziniu Li, Tian Xu, Zeyu Qin, Yang Yu, Zhi-Quan Luo
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 ISW-BC instead of BC (add reweighted supplementary data) | BC · MuJoCo, Noisy Expert, avg normalized | normalized score: 38% -> 69% | Improved |
| Add supplementary data naively (NBCU) vs BC under large shift | BC · MuJoCo, Noisy Expert, avg normalized | normalized score: 38% -> 35% | Hurt |
| Add supplementary data naively (NBCU) vs BC under small shift | BC · MuJoCo, Full Replay, avg normalized | normalized score: 38% -> 92% | Improved |
| ISW-BC instead of NBCU under large shift | NBCU · MuJoCo, Noisy Expert, avg normalized | normalized score: 35% -> 69% | Improved |
| ISW-BC instead of NBCU under small shift | NBCU · MuJoCo, Full Replay, avg normalized | normalized score: 92% -> 94% | No clear effect |
| ISW-BC instead of DWBC under small shift | DWBC · MuJoCo, Full Replay, avg normalized | normalized score: 44% -> 94% | Improved |
| ISW-BC instead of DemoDICE under large shift | DemoDICE · MuJoCo, Noisy Expert, avg normalized | normalized score: 49% -> 69% | Improved |
| ISW-BC instead of DWBC under large shift | DWBC · MuJoCo, Noisy Expert, avg normalized | normalized score: 62% -> 69% | Improved |
| Use ISW-BC instead of BC on Atari with noisy supplementary data | BC · Atari, Noisy Expert, avg normalized | normalized score: 32% -> 36% | Improved |
| NBCU instead of BC on Atari with noisy supplementary data | BC · Atari, Noisy Expert, avg normalized | normalized score: 32% -> 27% | Hurt |
| NBCU instead of BC on Atari with full replay | BC · Atari, Full Replay, avg normalized | normalized score: 32% -> 50% | Improved |
| ISW-BC instead of NBCU on Atari full replay | NBCU · Atari, Full Replay, avg normalized | normalized score: 50% -> 50% | No clear effect |
Closest papers by meaning
- How to Leverage Diverse Demonstrations in Offline Imitation Learning (ICML 2024 · similarity 0.92)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.91)
- Mitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution Correction (NeurIPS 2024 · similarity 0.90)
- Limited Preference Aided Imitation Learning from Imperfect Demonstrations (ICML 2024 · similarity 0.90)
- When Does Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning (NeurIPS 2026 · similarity 0.89)
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning (NeurIPS 2025 · similarity 0.89)
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.
