BATCH: Behavior Alignment for Cross-Embodiment Learning from Humans
BATCH: Behavior Alignment for Cross-Embodiment Learning from Humans
This research presents BATCH, a framework that aligns human and robot behaviors using self-supervised temporal representations and cycle-consistency objectives. It employs an autoregressive video-action diffusion model with behavior-aligned co-training and human-to-robot style transfer to improve generalization on unseen task configurations in real-world manipulation tasks compared to unaligned data and prior baselines.
The paper introduces BATCH, a framework that uses cycle-consistency and autoregressive video-action diffusion to align human demonstrations with robot behaviors for cross-embodiment transfer, reporting improved generalization on real-world manipulation tasks compared to unaligned data and prior baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.10148v1 Announce Type: new Abstract: Human demonstrations provide a scalable source of diverse manipulation data, but leveraging them for robot learning remains challenging due to differences in embodiment, viewpoint, action spaces, and temporal structure. We introduce BATCH, a cross-embodiment learning framework that explicitly aligns human and robot behaviors. BATCH learns self-supervised temporal repres
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org