Probe to Act: Elevating Browser-Use Agent via Active Visual Probing
Browser-use agents require seamless alignment between structured web metadata and visual information, while preserving relevant context across long interactions. Existing interfaces often rely on either screenshot-level action prediction or static Set-of-Marks overlays, leaving the model to resolve dense DOM-pixel alig...
Ke-Liang Li, Heng Wang, Chen Hu et al.
· 0 citations