aiqbench analysis
This research introduces KAN-We-Flow, a lightweight robotic policy leveraging Kolmogorov-Arnold Networks (KAN) and RWKV architectures to overcome the inference inefficiency of traditional diffusion models in 3D manipulation tasks. By replacing heavy UNet backbones with an efficient RWKV-KAN block and implementing Action Consistency Regularization, the method reduces parameter count by over 86% while maintaining state-of-the-art performance on standard benchmarks like Adroit and Meta-World. The approach demonstrates a viable path toward deploying high-fidelity visuomotor policies on resource-constrained hardware without sacrificing accuracy or speed.
KAN We Flow? Advancing Robotic Manipulation with 3D Flow Matching via KAN & RWKV
arXiv:2602.01115v3 Announce Type: replace Abstract: Diffusion-based visuomotor policies excel at modeling action distributions but are inference-inefficient, since recursively denoising from noise to policy requires many steps and heavy UNet backbones, which hinders deployment on resource-constrained robots. Flow matching alleviates the sampling burden by learning a one-step vector field, yet prior implementations still inherit large UNet-style architectures. In this work, we present KAN-We-Flow, a flow-matching policy that draws on recent advances in Receptance Weighted Key Value (RWKV) and Kolmogorov-Arnold Networks (KAN) from vision to build a lightweight and highly expressive backbone for 3D manipulation. Concretely, we introduce an RWKV-KAN block: an RWKV first performs efficient time/channel mixing to propagate task context, and a subsequent GroupKAN layer applies learnable spline-based, groupwise functional mappings to perform feature-wise nonlinear calibration of the action mapping on RWKV outputs. Moreover, we introduce an Action Consistency Regularization (ACR), a lightweight auxiliary loss that enforces alignment between predicted action trajectories and expert demonstrations via Euler extrapolation, providing additional supervision to stabilize training and improve policy precision. Without resorting to large UNets, our design reduces parameters by 86.8\%, maintains fast runtime, and achieves state-of-the-art success rates on Adroit, Meta-World, and DexArt benchmarks. Our project page can be viewed in \href{https://zhihaochen-2003.github.io/KAN-We-Flow.github.io/}{\textcolor{red}{link}}
Title:KAN We Flow? Advancing Robotic Manipulation with 3D Flow Matching via KAN & RWKV
Abstract:Diffusion-based visuomotor policies excel at modeling action distributions but are inference-inefficient, since recursively denoising from noise to policy requires many steps and heavy UNet backbones, which hinders deployment on resource-constrained robots. Flow matching alleviates the sampling burden by learning a one-step vector field, yet prior implementations still inherit large UNet-style architectures. In this work, we present KAN-We-Flow, a flow-matching policy that draws on recent advances in Receptance Weighted Key Value (RWKV) and Kolmogorov-Arnold Networks (KAN) from vision to build a lightweight and highly expressive backbone for 3D manipulation. Concretely, we introduce an RWKV-KAN block: an RWKV first performs efficient time/channel mixing to propagate task context, and a subsequent GroupKAN layer applies learnable spline-based, groupwise functional mappings to perform feature-wise nonlinear calibration of the action mapping on RWKV outputs. Moreover, we introduce an Action Consistency Regularization (ACR), a lightweight auxiliary loss that enforces alignment between predicted action trajectories and expert demonstrations via Euler extrapolation, providing additional supervision to stabilize training and improve policy precision. Without resorting to large UNets, our design reduces parameters by 86.8\%, maintains fast runtime, and achieves state-of-the-art success rates on Adroit, Meta-World, and DexArt benchmarks. Our project page can be viewed in \href{this https URL}{\textcolor{red}{link}}
Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2602.01115 [cs.RO]
(or arXiv:2602.01115v3 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2602.01115 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 309 words
Related coverage
Physical Time as a Missing Dimension of Robot Policy Memory Dynamics
arXiv · Oct 15, 2026
Distillation for Efficient Multitask Manipulation Policies via Conditional Flow Matching
arXiv · Sep 24, 2026
ActionMap: Robot Policy Learning via Voxel Action Heatmap
arXiv · Jun 11, 2026
ActionMap: Robot Policy Learning via Voxel Action Heatmap
arXiv · Jun 8, 2026