# XS-VLA: Teaching Tiny Vision-Language-Action Models with Spatial Supervision and Demonstration Conditioning

DevFeed: [XS-VLA: Teaching Tiny Vision-Language-Action Models with Spatial Supervision and Demonstration Conditioning](<https://devfeed.tech/articles/xs-vla-teaching-tiny-vision-language-action-models-with-spatial-supervision-and-demonstration-conditioning-128555.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2607.04171>)

Author: Iok Tong Lei, Ying Jie Yap, Wei Huang, Qingchen Xie, Qianzhi Li, Yujie Zhang, Xiaolong Liu, Zhidong Deng

Published: 2026-11-11T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Robotics](<https://devfeed.tech/topics/robotics.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [smolvla](<https://devfeed.tech/topics/smolvla.md>)

Tags: [control](<https://devfeed.tech/tags/control.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [distillation](<https://devfeed.tech/tags/distillation.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [kl-regularization](<https://devfeed.tech/tags/kl-regularization.md>), [learning](<https://devfeed.tech/tags/learning.md>), [robot](<https://devfeed.tech/tags/robot.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [supervision](<https://devfeed.tech/tags/supervision.md>), [training](<https://devfeed.tech/tags/training.md>), [vla](<https://devfeed.tech/tags/vla.md>)

## AI overview

XS-VLA uses a two-stage training framework to improve compact robot policies: Coarse-Grained Spatial Distillation initializes the backbone with teacher-derived region labels, then Latent Flow Matching trains an action-space velocity field conditioned on a demonstration latent, with KL regularization. The deployed 243.99M-parameter policy does not require the teacher or posterior encoder. In the reported results, it reaches 90.25% average LIBERO success in each of two seeds, versus 86.00% for a SmolVLA-0.25B baseline trained under the same settings. Across three Mobile ALOHA tasks, average strict success rises from 21.7% to 65.0%. Ablations include matched image pretraining and Huber/MSE controls.

## Source excerpt

arXiv:2607.04171v5 Announce Type: replace-cross Abstract: How can richer training supervision improve robot control while keeping the deployed policy compact? We present XS-VLA, a staged training framework using teacher-derived spatial labels and demonstration-conditioned action learning. Coarse-Grained Spatial Distillation (CSD) initializes the backbone through an auxiliary region-label task. Latent Flow Matching (LFM) then conditions an action-space velocity field on a demonstration latent, using KL regularization while jointly optimizing the backbone and action modules. The deployed policy contains 243.99M parameters and operates without the teacher or posterior encoder. XS-VLA achieves 90.25% average LIBERO success in each of two training seeds, compared with 86.00% for a SmolVLA-0.25B base trained under our settings. Ablations examine both training stages through matched image pretraining and Huber/MSE controls. On three Mobile ALOHA tasks, average strict success increases from 21.7% to 65.0%. These results demonstrate the control utility of auxiliary representation initialization and regularized demonstration-conditioned flow learning for compact VLA policies