ViRA: visual agents that zoom into high-resolution photos, trained with CrossRPO on contrast sets. Models, benchmark and training data.
-
ViRA: Toward Ultra-High-Resolution Visual Agents via Cross-Input Relative Policy Optimization
Paper • 2611.21605 • Published • 4 -
huaweilin/ViRA-9B
Image-Text-to-Text • 9B • Updated • 121 • 2 -
huaweilin/ViRA-27B
Image-Text-to-Text • 27B • Updated • 120 • 3 -
huaweilin/ViRA-Bench
Viewer • Updated • 840 • 350 • 2