List your company

Multimodal Question Answering on ScienceQA (SQA) (test)

70.2SQA AccuracyVanilla (LLaVA-NeXT-7B)
66.87267.73668.669.464Jan 30, 2026

Evaluation results

MethodLinks
Vanilla (LLaVA-NeXT-7B)#Tokens=28802026.0170.2——9.6—1,512.1—
VisionTrim#Tokens=3202026.0169.6—61.40.891.7101.893.3
VisionZip#Tokens=3202026.0167.3—32.71.683.3180.488.1
SparseVLM#Tokens=3202026.0167—30.61.584.416888.9