LightNav-0 GitHub repo: Qwen3-VL-based generalist navigation model tops 700 stars
rsasaki0109 · x · 2026-10-14
The follow-up tweet links the LightNav-0 GitHub repo: a compact generalist embodied navigation model that elicits spatial intelligence from a pretrained Qwen3-VL and aligns it with navigation, without task-specific prediction heads.
Diverse tasks share one token interface, with dual-channel pointing expressing task- and scene-level information. The repo demo shows humanoid, quadruped, wheeled, and aerial robots autonomously following language-named targets in an unseen park, no teleoperation. Apache-2.0 licensed, with modules for docker, evtbench, habitatserver, mujocodemo, and robotdeploy; 706 stars, 53 forks.
Related event: LightNav-0 Open-Sourced: Qwen3-VL-Powered Navigation for Four Robot Types(2 posts)→
More from Embodied
- Physical Intelligence opens Pi robotics model API beta, already in 150 factories — svlevine · 2026-10-14
- The 'sophisticated device' in spy case is a $14.90 Seeed Studio ESP32 hobby kit — wavefnx · 2026-10-14
- KILVO: Fusing Encoders, IMU, LiDAR and Vision for Humanoid Robot Odometry — rsasaki0109 · 2026-10-14
- KILVO: robust multimodal odometry for humanoid robots fusing encoders, IMU, LiDAR and vision — rsasaki0109 · 2026-10-14
- Robotics firm Node is hiring a full-time robot dog walker in Palo Alto — ashleyyjoelle · 2026-10-14
- homeview wall-mounted smart display wins fans: family calendar, lists, zero ads — ow · 2026-10-14