Research map / Reinforcement learning / Imitation learning
When Does Predictive Inverse Dynamics Outperform Behavior Cloning?
Lukas Schäfer, Pallavi Choudhury, Abdelhak Lemkhenter, Chris Lovett, Somjit Nath, Luis França, Matheus Ribeiro Furtado de Mendonça, Alex Lamb, Riashat Islam, Siddhartha Sen, John Langford, Katja Hofmann and 1 more
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 PIDM (future-state conditioning) instead of BC | offline imitation learning · 2D navigation (human demos) | efficiency ratio eta_PIDM(90%): 3.5 | Improved |
| use PIDM instead of BC at 80% success threshold | offline imitation learning · 3D Bleeding Edge Tour | eta_PIDM(80%): 1.66 | Improved |
| collect training data with deterministic A* planner instead of human demonstrations (ablation) | PIDM · 2D Multiroom | demonstrations needed to reach 90% performance: 5x fewer demonstrations | Improved |
| lower state prediction error (less diverse data) (ablation) | PIDM · 2D navigation | rollout performance | Improved |
| use PIDM vs BC with 5 demonstrations | PIDM · 3D Bleeding Edge Tour | success rate | No clear effect |
| use PIDM over BC with 15 demos vs BC with 25 demos | PIDM · 3D Bleeding Edge Tour | success rate: 81% (BC, 25 demos) vs 87% (PIDM, 15 demos) | Improved |
| use PIDM instead of BC on max goal ratio in Maze | PIDM · 2D navigation: Maze | max goal ratio: 0.95 -> 0.99 | Improved |
| use PIDM instead of BC on max goal ratio in Multiroom | PIDM · 2D navigation: Multiroom | max goal ratio: 0.99 -> 0.98 | No clear effect |
Closest papers by meaning
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.88)
- Mitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution Correction (NeurIPS 2024 · similarity 0.87)
- Diffusion Model-Augmented Behavioral Cloning (ICML 2024 · similarity 0.86)
- How to Leverage Diverse Demonstrations in Offline Imitation Learning (ICML 2024 · similarity 0.86)
- Robust Visual Imitation Learning with Inverse Dynamics Representations (AAAI 2024 · similarity 0.85)
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning (NeurIPS 2025 · 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.
