List your company
ResearchFormerly WizWand

Multimodal Understanding on MME (Score, Speed-up)

2,388MME ScoreVanilla
1,022.44881,376.96691,731.4852,086.0031Apr 10, 2026Apr 15, 2026Apr 20, 2026Apr 26, 2026May 1, 2026May 6, 2026May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
VanillaModels=BAGEL-7B, Avg. Tokens=100%2026.052,388—100
VSCANModels=BAGEL-7B, Avg. Tokens=50%2026.052,339—94.4
IVC-PruneModels=BAGEL-7B, Avg. Tokens=50%2026.052,332—98.6
G2TRModels=BAGEL-7B, Avg. Tokens=50%2026.052,318—99
PDropModels=BAGEL-7B, Avg. Tokens=61%2026.052,312—94.1
FastVModels=BAGEL-7B, Avg. Tokens=54%2026.052,306—95.8
W-FastVModels=BAGEL-7B, Avg. Tokens=54%2026.052,287—95.6
VanillaModels=InternVL-U, Avg. Tokens=100%2026.052,057—100
FastVModels=InternVL-U, Avg. Tokens=54%2026.052,037—92.9
G2TRModels=InternVL-U, Avg. Tokens=50%2026.052,034—94.9
IVC-PruneModels=InternVL-U, Avg. Tokens=50%2026.052,020—93.7
VSCANModels=InternVL-U, Avg. Tokens=50%2026.051,968—92.2
PDropModels=InternVL-U, Avg. Tokens=61%2026.051,953—92
W-FastVModels=InternVL-U, Avg. Tokens=54%2026.051,926—92.3
AHD (Ours)2026.041,075.7215.98—
MMaDA2026.041,074.971—