ArcBench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ArcBench
100Win Rate
54
ArcBench VLM-as-Judge Closed
58.5TQ
15
ArcBench VLM-as-Judge Open
86.11TQ
15
ArcBench VLM-based Q/A Quiz Closed
97.6Story Accuracy
15
ArcBench VLM-based Q/A Quiz Open
95.69Accuracy (Story)
15