Reproduce “From Pixels to Actions - Towards Native Vision-Language-Action Policies” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weichen Fan, Haiwen Diao, Penghao Wu, Siyuan Hu, Ziqi Huang, Tianqi Liu, Mingze Sun, Ziyi Wang and 9 more
NEO-VLA combines image patches and language in one backbone and lets an action expert read selected layers to generate continuous action chunks. In frozen-backbone LIBERO evaluation, it reaches 90.9% average success, versus 79.5% for Qwen3-VL 2B. The comparison does not isolate encoder removal as the cause, and performance under object-position perturbations falls to 3.0%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.