List your company
ResearchFormerly WizWand

Multimodal Benchmarking on MMB

100Average PerformanceLLaVA-NeXT-7B
30.413648.479366.54584.6107Mar 1, 2026Mar 6, 2026Mar 12, 2026Mar 18, 2026Mar 23, 2026Mar 29, 2026Apr 4, 2026
Updated 6mo ago

Evaluation results

MethodLinks
LLaVA-NeXT-7BBase Model=LLaVA-NeXT-7B, Retained Tokens=28802026.0310065.8——
VisionZipBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0399.865.8——
AgilePrunerBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0399.6465.9——
DivPruneBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0398.965.8——
SparseVLMBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0398.865.9——
PruMerge+Base Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0398.364.6——
AgilePrunerBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0397.9464.5——
PDropBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.039764.1——
DivPruneBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.039763.9——
VisionZipBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0396.463.3——
PruMerge+Base Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0395.763——
FastVBase Model=LLaVA-NeXT-7B, Retained Tokens=6402026.0395.263.1——
SparseVLMBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0394.663.1——
PDropBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0382.555.5——
FastVBase Model=LLaVA-NeXT-7B, Retained Tokens=3202026.0379.953.4——
TPRLBackbone=LLaVA-1.5-VL-13B, Token=192*2026.0368.95—100.5—
LLaVA-1.5-VL-13BBackbone=LLaVA-1.5-VL-13B, Token=5762026.0368.64—100—
TPRLBackbone=LLaVA-1.5-VL-13B, Token=128*2026.0367.5—98.3—
FastVBackbone=LLaVA-1.5-VL-13B, Token=1922026.0364.8—94.5—
VanillaBackbone=LLaVA-1.5-7B, Retained Tokens=5762026.0464.3100—100
VisionZipBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0464.198.75—99.69
RCPBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0463.6199.06—98.93
DARTBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0463.2198.57—98.3
PDropBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0462.8997.07—97.8
HiREDBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0462.4196.53—97.06
FastVBackbone=LLaVA-1.5-VL-13B, Token=1282026.0361.2—89.2—
FastVBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0461.288.71—95.18
HiREDBackbone=LLaVA-1.5-7B, Retained Tokens=1282026.0461.1295.2—95.05
PDropBackbone=LLaVA-1.5-7B, Retained Tokens=1282026.0460.895.08—94.55
DARTBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0460.2394.4—93.66
ToMeBackbone=LLaVA-1.5-7B, Retained Tokens=1922026.0460.1388.32—93.51
VisionZipBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0460.194.81—93.47
HiREDBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0459.8391.46—93.04
RCPBackbone=LLaVA-1.5-7B, Retained Tokens=1282026.0459.3496.73—92.29
RCPBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0459.2995—92.21
FastVBackbone=LLaVA-1.5-7B, Retained Tokens=1282026.0456.182.86—87.25
ToMeBackbone=LLaVA-1.5-7B, Retained Tokens=1282026.0452.9782.75—82.38
FastVBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0447.773.88—74.19
ToMeBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0443.4373.34—67.54
PDropBackbone=LLaVA-1.5-7B, Retained Tokens=642026.0433.0975.89—51.47
ID-SelectionImportance Criterion=LLM second-layer cross-modal attention, Retain Tokens=642026.04—61.8——
ID-SelectionImportance Criterion=LLM second-layer cross-modal attention, Retain Tokens=322026.04—60.9——
ID-SelectionImportance Criterion=unified score, Retain Tokens=162026.04—58.8——
LLaVA-1.5-7BRetain Tokens=5762026.04—64.3——