LLaVA Evaluation Suite
Benchmarks
TaskDatasetSOTA resultTrendResults
LLaVA Evaluation Suite GQA, MMB, MMMU, MME, VQAv2, VQAtext, POPE, SQA, SEED, MMVet 1.5 7B (various)
62GQA
46
LLaVA Evaluation Suite 1.5 (test val)
63.5GQA
19
LLaVA-1.5 Evaluation Suite OKVQA
78.52VQAv2 Accuracy
6