Vero is an open VLM family trained via RL on Vero-600K (600K samples from 59 datasets across six categories) with task-routed rewards, achieving SOTA gains of 3.6-5.3 points on 30 visual reasoning benchmarks.
Spatial & Action receives the largest share (0.273) due to its low initial accuracy, while STEM receives the smallest (0.170)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Vero: An Open RL Recipe for General Visual Reasoning
Vero is an open VLM family trained via RL on Vero-600K (600K samples from 59 datasets across six categories) with task-routed rewards, achieving SOTA gains of 3.6-5.3 points on 30 visual reasoning benchmarks.