KAN We Flow? Advancing Robotic Manipulation with 3D Flow Matching via KAN & RWKV
Zhihao Chen, Yiyuan Ge, Ziyang Wang · 2026
Read KAN We Flow? Advancing Robotic Manipulation with 3D Flow Matching via KAN & RWKV by Zhihao Chen, Yiyuan Ge, Ziyang Wang (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-2602-01115.txt.
Description
Zhihao Chen and co-authors build KAN-We-Flow, a light flow plan for 3D robot hand work. Diffusion hand plans model moves well but crawl, since they clean noise step by step through huge nets, barring use on small bots. Their RWKV-KAN block first mixes time and channel cues with little cost, then bends each trait with learnable spline maps. A spare loss lines up planned moves with expert moves, and the plan cut weights by 86.8% while topping three test sets.
Catalog record
| CREATOR | Zhihao Chen, Yiyuan Ge, Ziyang Wang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37600 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01115