SalArt-VQA Leaderboard
Accuracy (%), higher is better.
Loading results…
| Model | Artifact · 475 | Ref-Real · 356 | Ref-PG · 119 | Overall | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
All Q1–4 / Q2–4: all indicated answers on the same image must be correct. Ref-Real: real references. Ref-PG: generated references (Q2–4 only).
Overall Q1 and All Q1–4 use 831 images; All Q2–4 uses 950. Bold marks the best score among the models shown.