ViRA trains high-resolution visual agents with CrossRPO contrast sets
Authors report 63.3% ViRA-Bench set accuracy versus 59.3% for GRPO after about 3,500 training items.
48
Authors report 63.3% ViRA-Bench set accuracy versus 59.3% for GRPO after about 3,500 training items.
ViRA-9B and ViRA-27B, trained on about 3,500 items each, are reported ahead of GRPO and several closed models.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 980 readers · 1.1K views · 92 reads · all time 211K views