Research map / Reinforcement learning / Imitation learning
Koopman-Assisted Trajectory Synthesis: A Data Augmentation Framework for Offline Imitation Learning
Jin Wang, Pengcheng He, Ke Jiang, Xiaoyang Tan
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 |
|---|---|---|---|
| add inverse dynamics model for action inference (ablation) | KATS · antmaze-umaze | normalized score: 78.4 ± 4.8 -> 96.9 ± 0.8 | Improved |
| add inverse dynamics model for action inference (ablation) | KATS · maze2d-large | normalized score: 41.5 ± 15.8 -> 100.1 ± 7.2 | Improved |
| use learnable sigma network instead of fixed analytical matrix (ablation) | KATS · antmaze-umaze | normalized score: 87.3 ± 2.6 -> 96.9 ± 0.8 | Improved |
| use learnable sigma network instead of fixed analytical matrix (ablation) | KATS · maze2d-large | normalized score: 89.2 ± 12.2 -> 100.1 ± 7.2 | Improved |
| use state-equivariant decoupled design instead of action-coupled KFC (ablation) | KATS+BC · maze2d-large | normalized score: 55.9 ± 9.1 -> 100.1 ± 7.2 | Improved |
| add KATS augmentation to BC | BC · antmaze-umaze | normalized score: 74.0 ± 1.2 -> 96.9 ± 0.8 | Improved |
| replace KFC augmentation with KATS | BC · antmaze-umaze | normalized score: 79.1 ± 3.4 -> 96.9 ± 0.8 | Improved |
| replace KFC augmentation with KATS | BC · hopper-medium-expert | normalized score: 70.2 ± 1.3 -> 112.7 ± 3.6 | Improved |
| apply KATS augmentation to BC and SRA datasets | BC, SRA · D4RL offline IL | performance | Improved |
| KATS+BC vs offline RL baselines in few-shot | KATS+BC · Antmaze-u few-shot (10%) | normalized score: 83.1±7.2 vs TELS 88.7±7.7 | Mixed |
Closest papers by meaning
- Imitation from Observations with Trajectory-Level Generative Embeddings (NeurIPS 2026 · similarity 0.89)
- 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 0.88)
- How to Leverage Diverse Demonstrations in Offline Imitation Learning (ICML 2024 · similarity 0.88)
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning (NeurIPS 2025 · similarity 0.88)
- CCIL: Continuity-Based Data Augmentation for Corrective Imitation Learning (ICLR 2024 · similarity 0.88)
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.
