FocusVLA routes robot camera views by predicted interaction intent across task stages
iscas-cloud · hf · 2026-10-26
A paper from the Chinese Academy of Sciences' Institute of Software and collaborators, FocusVLA improves Vision-Language-Action models for robotic manipulation via future-intent guided dynamic view routing:
- Motivation: different manipulation stages (approach, grasp, transfer, place) need different visual focus; statically relying on either global or local views loses key context
- Method: dynamically routes between global and local camera views based on predicted interaction intent at each task stage, with dynamic routing weights evolving over time
- Results: charts show success-rate gains across short/medium/long horizons, reaching about 85.9% on the long horizon
More from Embodied
- AI researchers guessed robots wouldn't outrun humans until 2034 — it already happened — ben_j_todd · 2026-10-26
- Shanghai AI Lab-led team releases 100-hour egocentric wearable dataset for humanoid training — ChuckDBrooks · 2026-10-26
- RealSense D585 Pro AI-Powered 3D Stereo Depth Camera Unboxed for Physical AI — chrismatthieu · 2026-10-26
- Physical world agent swarm demo shared on X — pduan · 2026-10-26
- Scoble shares his full list of AI wearable recording mics — best ones 'yet to come' — Scobleizer · 2026-10-26
- 'Lamp Swarm': a desk of lamp-shaped robots shows what physical AI work looks like — dee_hw · 2026-10-26