LightZeroNav: Zero-Shot Vision Language Navigation in Continuous Environments Based on Lightweight VLMs

In three linesLightZeroNav addresses zero-shot vision-language navigation in continuous environments using lightweight VLMs. Built on Qwen3-VL-8B, it handles information redundancy, noisy progress estimation, and task entanglement. Achieves GPT-4o-level performance without task-specific training, graph search, or waypoint predictors.

Summary generated by Claude — human-verified