Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• SaPaVe achieves up to 31.25% higher success rates in real-world tasks compared to recent vision-language-action models like GR00T N1 and π₀.result 90%
• SaPaVe decouples camera and manipulation actions rather than placing them in a shared action space.method 85%
• SaPaVe follows a bottom-up training strategy: first training semantic camera control on large-scale data, then jointly optimizing both action types using hybrid data.method 85%
• The ActiveViewPose-200K dataset contains 200k image-language-camera movement pairs for semantic camera movement learning.technical 90%
• SaPaVe includes a 3D geometry-aware module that improves execution robustness under dynamic viewpoints.technical 85%
• ActiveManip-Bench is the first benchmark for evaluating active manipulation beyond fixed-view settings.technical 80%
• Existing methods struggle to unify semantic-driven active perception with robust, viewpoint-invariant execution.limitation 75%
• Tightly coupled perception and execution, when trained with decoupled yet coordinated strategies, enable efficient and generalizable active manipulation.result 70%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card