Offline Discovery of Interpretable Skills from Multi-Task Trajectories
Chongyu Zhu, Mithun Vanniasinghe, Jiayu Chen, Chi-Guhn Lee · 2026
Read Offline Discovery of Interpretable Skills from Multi-Task Trajectories by Chongyu Zhu, Mithun Vanniasinghe, Jiayu Chen, Chi-Guhn Lee (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-01018.txt.
Description
Chongyu Zhu and co-authors teach robots to find reusable skills from long mixed-task demo tapes with no rewards or marks. A first pass cuts tapes into rough chunks with a codebook net steered by weak tags. A second pass sharpens those cuts before grouping locks in skill edges. A third pass builds a layered control plan on those skills, with a learned stop rule. The skills read as real actions to people, and the robot strung them in a new order for a fresh task.
Catalog record
| CREATOR | Chongyu Zhu, Mithun Vanniasinghe, Jiayu Chen, Chi-Guhn Lee |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40213 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01018