Xiao-Ming Wu

VLANeXt systematically studies what actually matters for VLA performance: policy heads, meta-queries, action chunking, flow matching, temporal history, multi-view, and proprioception, and shows that a simple time-series forecasting perspective improves action prediction. 📄 Paper: arxiv.org/abs/2602.18532 💻 Code: github.com/DravenALG/VLAN… If you’re working on VLAs, also check out this curated list of resources: 🔎 github.com/DravenALG/awes…

VLANeXt: Recipes for Building Strong VLA Models A systematic study distilling 12 key findings for Vision-Language-Action models that outperforms prior SOTA on LIBERO benchmarks with a unified, easy-to-use codebase.

Thanks @_akhaliq for sharing our work! Our codebase can be found at github.com/DravenALG/VLAN…, which can enable the full combination of the 12 design spaces explored in our paper. Welcome to try!

@Rainmaker1973 Very interesting!

Human history through selfies (with AI) [📹 edmundsplanet: buff.ly/3oQh6pF] twitter.com/AliciaMimundo/…