See What Matters: Differentiable Grid Sample Pruning for Generalizable Vision-Language-Action Model

Yixu Feng, Zinan Zhao, Yanxiang Ma, Chenghao Xia, Chengbin Du, Yunke Wang, Chang Xu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:30552-30578, 2026.

Abstract

Vision-Language-Action (VLA) models have shown remarkable promise in robotics manipulation, yet their high computational cost hinders real-time deployment. Existing token pruning methods suffer from a fundamental trade-off: aggressive compression using pruning inevitably discards critical geometric details like contact points, leading to severe performance degradation. This forces a compromise, limiting the achievable compression rate and thus the potential speedup. We argue that breaking this trade-off requires rethinking compression as a geometry-aware, continuous token resampling in the vision encoder. To this end, we propose the Differentiable Grid Sampler (GridS), a plug-and-play module that performs task-aware, continuous resampling of visual tokens in VLA. By adaptively predicting a minimal set of salient coordinates and extracting features via differentiable interpolation, GridS preserves essential spatial information while achieving drastic compression (with fewer than 10% original visual tokens). Experiments on both LIBERO benchmark and a real robotic platform demonstrate that validating the lowest feasible visual token count reported to date, GridS achieves a 76% reduction in FLOPs with no degradation in the success rate. The code is available at https://github.com/Fediory/Grid-Sampler.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-feng26x, title = {See What Matters: Differentiable Grid Sample Pruning for Generalizable Vision-Language-Action Model}, author = {Feng, Yixu and Zhao, Zinan and Ma, Yanxiang and Xia, Chenghao and Du, Chengbin and Wang, Yunke and Xu, Chang}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {30552--30578}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/feng26x/feng26x.pdf}, url = {https://proceedings.mlr.press/v306/feng26x.html}, abstract = {Vision-Language-Action (VLA) models have shown remarkable promise in robotics manipulation, yet their high computational cost hinders real-time deployment. Existing token pruning methods suffer from a fundamental trade-off: aggressive compression using pruning inevitably discards critical geometric details like contact points, leading to severe performance degradation. This forces a compromise, limiting the achievable compression rate and thus the potential speedup. We argue that breaking this trade-off requires rethinking compression as a geometry-aware, continuous token resampling in the vision encoder. To this end, we propose the Differentiable Grid Sampler (GridS), a plug-and-play module that performs task-aware, continuous resampling of visual tokens in VLA. By adaptively predicting a minimal set of salient coordinates and extracting features via differentiable interpolation, GridS preserves essential spatial information while achieving drastic compression (with fewer than 10% original visual tokens). Experiments on both LIBERO benchmark and a real robotic platform demonstrate that validating the lowest feasible visual token count reported to date, GridS achieves a 76% reduction in FLOPs with no degradation in the success rate. The code is available at https://github.com/Fediory/Grid-Sampler.} }
Endnote
%0 Conference Paper %T See What Matters: Differentiable Grid Sample Pruning for Generalizable Vision-Language-Action Model %A Yixu Feng %A Zinan Zhao %A Yanxiang Ma %A Chenghao Xia %A Chengbin Du %A Yunke Wang %A Chang Xu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-feng26x %I PMLR %P 30552--30578 %U https://proceedings.mlr.press/v306/feng26x.html %V 306 %X Vision-Language-Action (VLA) models have shown remarkable promise in robotics manipulation, yet their high computational cost hinders real-time deployment. Existing token pruning methods suffer from a fundamental trade-off: aggressive compression using pruning inevitably discards critical geometric details like contact points, leading to severe performance degradation. This forces a compromise, limiting the achievable compression rate and thus the potential speedup. We argue that breaking this trade-off requires rethinking compression as a geometry-aware, continuous token resampling in the vision encoder. To this end, we propose the Differentiable Grid Sampler (GridS), a plug-and-play module that performs task-aware, continuous resampling of visual tokens in VLA. By adaptively predicting a minimal set of salient coordinates and extracting features via differentiable interpolation, GridS preserves essential spatial information while achieving drastic compression (with fewer than 10% original visual tokens). Experiments on both LIBERO benchmark and a real robotic platform demonstrate that validating the lowest feasible visual token count reported to date, GridS achieves a 76% reduction in FLOPs with no degradation in the success rate. The code is available at https://github.com/Fediory/Grid-Sampler.
APA
Feng, Y., Zhao, Z., Ma, Y., Xia, C., Du, C., Wang, Y. & Xu, C.. (2026). See What Matters: Differentiable Grid Sample Pruning for Generalizable Vision-Language-Action Model. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:30552-30578 Available from https://proceedings.mlr.press/v306/feng26x.html.

Related Material