robotics
The Robot Looks Through the Camera Its Next Move Needs
Future interaction intent routes attention between global and wrist views instead of letting one camera dominate every manipulation stage.
Summary
Future interaction intent routes attention between global and wrist views instead of letting one camera dominate every manipulation stage.
Multi-view robot policies receive complementary global and local observations, yet the authors found persistent over-reliance on the wrist camera. FocusVLA predicts future interaction intent from action history and uses that estimate to route visual information differently across task stages. The modules can be attached to existing vision-language-action frameworks without changing their cores. On RoboTwin 2.0, the paper reports a 10.10-point average success-rate improvement over the strongest pi-0.5 baseline; in real-world tests, success rose from 71.7 to 85.0 percent. These figures describe the selected benchmark, baseline and physical tasks and do not demonstrate that predicted intent will choose the right view under every occlusion or distribution shift.
Why it matters
Future interaction intent routes attention between global and wrist views instead of letting one camera dominate every manipulation stage.
Limits and context
- These figures describe the selected benchmark, baseline and physical tasks and do not demonstrate that predicted intent will choose the right view under every occlusion or distribution shift.
Key claims
Future interaction intent routes attention between global and wrist views instead of letting one camera dominate every manipulation stage.
Qualification: These figures describe the selected benchmark, baseline and physical tasks and do not demonstrate that predicted intent will choose the right view under every occlusion or distribution shift.
Evidence: source-2026-10-25-010
Sources
- arXiv preprint 2610.31423, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.