AI Paper Atlas

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

ICLR 2026 · Reinforcement learning › Imitation learning · approach: Offline imitation from imperfect data

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 changedWhereMeasuredEffect
add inverse dynamics model for action inference (ablation)KATS · antmaze-umazenormalized score: 78.4 ± 4.8 -> 96.9 ± 0.8Improved
add inverse dynamics model for action inference (ablation)KATS · maze2d-largenormalized score: 41.5 ± 15.8 -> 100.1 ± 7.2Improved
use learnable sigma network instead of fixed analytical matrix (ablation)KATS · antmaze-umazenormalized score: 87.3 ± 2.6 -> 96.9 ± 0.8Improved
use learnable sigma network instead of fixed analytical matrix (ablation)KATS · maze2d-largenormalized score: 89.2 ± 12.2 -> 100.1 ± 7.2Improved
use state-equivariant decoupled design instead of action-coupled KFC (ablation)KATS+BC · maze2d-largenormalized score: 55.9 ± 9.1 -> 100.1 ± 7.2Improved
add KATS augmentation to BCBC · antmaze-umazenormalized score: 74.0 ± 1.2 -> 96.9 ± 0.8Improved
replace KFC augmentation with KATSBC · antmaze-umazenormalized score: 79.1 ± 3.4 -> 96.9 ± 0.8Improved
replace KFC augmentation with KATSBC · hopper-medium-expertnormalized score: 70.2 ± 1.3 -> 112.7 ± 3.6Improved
apply KATS augmentation to BC and SRA datasetsBC, SRA · D4RL offline ILperformanceImproved
KATS+BC vs offline RL baselines in few-shotKATS+BC · Antmaze-u few-shot (10%)normalized score: 83.1±7.2 vs TELS 88.7±7.7Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy