Mixture of Horizons in Action Chunking

Dong Jing, Gang Wang, Jiaqi Liu, Weiliang Tang, Zelong Sun, Yunchao Yao, Zhenyu Wei, Yunhui Liu, Zhiwu Lu, Mingyu Ding
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:54350-54370, 2026.

Abstract

Vision-language-action models exhibit an inherent trade-off in action chunk length (“horizon”): longer horizons improve global foresight but degrade fine-grained local control, while shorter ones yield the opposite. To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. In brief, MoH rearranges the action chunk into several segments with different horizons, processes them in parallel with a shared action transformer, and fuses outputs with a light linear gate. It offers three appealing benefits. i) Long-term foresight and short-term precision are jointly exploited within a single model. ii) MoH is plug-and-play for full-attention action modules with minimal training or inference overhead. iii) MoH enables dynamic inference with adaptive horizons, which selects stable actions through cross-horizon consensus, achieving 2.5$\times$ higher throughput than baselines while preserving superior performance. Extensive experiments over flow-based and one-step regression policies demonstrate that MoH yields consistent and significant gains on both simulations and real-world tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-jing26f, title = {Mixture of Horizons in Action Chunking}, author = {Jing, Dong and Wang, Gang and Liu, Jiaqi and Tang, Weiliang and Sun, Zelong and Yao, Yunchao and Wei, Zhenyu and Liu, Yunhui and Lu, Zhiwu and Ding, Mingyu}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {54350--54370}, 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/jing26f/jing26f.pdf}, url = {https://proceedings.mlr.press/v306/jing26f.html}, abstract = {Vision-language-action models exhibit an inherent trade-off in action chunk length (“horizon”): longer horizons improve global foresight but degrade fine-grained local control, while shorter ones yield the opposite. To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. In brief, MoH rearranges the action chunk into several segments with different horizons, processes them in parallel with a shared action transformer, and fuses outputs with a light linear gate. It offers three appealing benefits. i) Long-term foresight and short-term precision are jointly exploited within a single model. ii) MoH is plug-and-play for full-attention action modules with minimal training or inference overhead. iii) MoH enables dynamic inference with adaptive horizons, which selects stable actions through cross-horizon consensus, achieving 2.5$\times$ higher throughput than baselines while preserving superior performance. Extensive experiments over flow-based and one-step regression policies demonstrate that MoH yields consistent and significant gains on both simulations and real-world tasks.} }
Endnote
%0 Conference Paper %T Mixture of Horizons in Action Chunking %A Dong Jing %A Gang Wang %A Jiaqi Liu %A Weiliang Tang %A Zelong Sun %A Yunchao Yao %A Zhenyu Wei %A Yunhui Liu %A Zhiwu Lu %A Mingyu Ding %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-jing26f %I PMLR %P 54350--54370 %U https://proceedings.mlr.press/v306/jing26f.html %V 306 %X Vision-language-action models exhibit an inherent trade-off in action chunk length (“horizon”): longer horizons improve global foresight but degrade fine-grained local control, while shorter ones yield the opposite. To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. In brief, MoH rearranges the action chunk into several segments with different horizons, processes them in parallel with a shared action transformer, and fuses outputs with a light linear gate. It offers three appealing benefits. i) Long-term foresight and short-term precision are jointly exploited within a single model. ii) MoH is plug-and-play for full-attention action modules with minimal training or inference overhead. iii) MoH enables dynamic inference with adaptive horizons, which selects stable actions through cross-horizon consensus, achieving 2.5$\times$ higher throughput than baselines while preserving superior performance. Extensive experiments over flow-based and one-step regression policies demonstrate that MoH yields consistent and significant gains on both simulations and real-world tasks.
APA
Jing, D., Wang, G., Liu, J., Tang, W., Sun, Z., Yao, Y., Wei, Z., Liu, Y., Lu, Z. & Ding, M.. (2026). Mixture of Horizons in Action Chunking. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:54350-54370 Available from https://proceedings.mlr.press/v306/jing26f.html.

Related Material