Reproduce “NereidVLA: Learning to Understand and Navigate the Underwater World” 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

Chenjie Zhao, Zhiliang Zhu, Quan Chen, Yuquan Xue, Xingxue Shi, Puyue Wang, Zhiqiang Zhao, Kangye Zheng and 3 more

NereidVLA connects marine visual knowledge to underwater robot navigation by first training Sea2Know on marine vision-language data without paired motion, then adapting it with simulated expert trajectories. Its policy predicts local 3D paths from image history and language for a feedback controller. In matched simulation, success rises from 72.4% to 84.4%; pool trials show 78.75%, not open-water generalization.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.