Reproduce “LightVLN: Efficient Aerial Vision-and-Language Navigation with Compact Memory and History-Guided Local Aggregation” 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

Yiming Zhao, Tianshun Li, Jingle He, Ruonan Chai, Xinhu Zheng

LightVLN compresses visual context for long-route drone navigation: one token per historical frame and 32 aggregated tokens instead of 256 for the current view. With a 0.5B language backbone and up to 16 history tokens, it uses at most 48 observation-derived tokens. It reaches 36.14% success on OpenFly Test-Unseen; campus success is 8.00% in real-to-simulation hardware-in-the-loop testing, not autonomous flight.

Summary by alphaXiv.

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