List your company

Multimodal Understanding on GQA, MMB, MMB-CN, MME, POPE, SQA, and TextVQA

63.1GQA AccuracyVanilla
52.755.458.160.8Jun 23, 2026

Evaluation results

MethodLinks
VanillaBackbone=InternVL2.5-8B, Token Retention Ratio=100%, Visual Token Count=1792 Tokens, Evaluation Protocol=Upper Bound2026.0663.184.782.62,33290.59876.8100
FastVBackbone=InternVL2.5-8B, Token Retention Ratio=33.3%, Evaluation Protocol=Retain Tokens2026.0661.183.381.52,25190.397.375.198.2
PriorTRBackbone=InternVL2.5-8B, Token Retention Ratio=33.3%, Evaluation Protocol=Retain Tokens2026.066183.883.12,27590.597.373.598.4
PriorTRBackbone=InternVL2.5-8B, Token Retention Ratio=22.2%, Evaluation Protocol=Retain Tokens2026.0660.282.781.42,26390.895.771.997.2
FastVBackbone=InternVL2.5-8B, Token Retention Ratio=22.2%, Evaluation Protocol=Retain Tokens2026.0659.68180.32,17689.694.973.396
PriorTRBackbone=InternVL2.5-8B, Token Retention Ratio=11.1%, Evaluation Protocol=Retain Tokens2026.0656.880.878.72,1508991.267.593.2
FastVBackbone=InternVL2.5-8B, Token Retention Ratio=11.1%, Evaluation Protocol=Retain Tokens2026.0653.17472.81,97084.78865.387.5