List your company

Multimodal Understanding on LLaVA Evaluation Suite 1.5

101.15Average ScoreMS-Resampler
89.55492.564595.57598.5855Aug 25, 2025Oct 15, 2025Dec 6, 2025Jan 26, 2026Mar 19, 2026May 9, 2026Jun 30, 2026

Evaluation results

MethodLinks
MS-ResamplerBackbone=LLaVA-1.5-7B, Vision token count=1442026.06101.156265.6—8771.278——33.1———65.2—————57.21,45333.9
TokenPackerBackbone=LLaVA-1.5-7B, Vision token count=1442026.06100.256265—8768.978——33———65.1—————57.11,45432.6
MS-ResamplerBackbone=LLaVA-1.5-7B, Vision token count=642026.06100.0762.164.5—87.670.678——30.8———66.1—————56.61,46932.9
VanillaToken Count=576, Reduction Ratio=0%2026.0110061.964.71,86285.969.578.558.258.631.150.1———————————
LLaVA-1.5-7B# Vision Tokens=5762026.0210061.9—1,86485.969.578.558.2——————————————
LLaVA-1.5-13BRetained Tokens=576, Pruning Rate=0.0%2025.0910063.267.71,81885.972.88061.366.9———36.4—————————
LLaVA-1.5 Vanilla 13BTokens Retained=576, Compression Ratio=100%2025.08100—————————————————————
VanillaLoc.=-, Retained Tokens=576, Pruning Ratio=0%2026.03100—————————————————————
VanillaEarly Cmp.=−, Retained Tokens=576, Compression Ratio=100%2026.03100—————————————————————
VanillaBase Model=LLaVA-1.5 13B, Retained Tokens=576 (100%)2026.0310063.267.71,81885.972.88061.3——56.6—36.466.9————————
VanillaBackbone=LLaVA-1.5-7B, Vision token count=1442026.061006264.3—85.96878.5——31.6———66.1—————58.21,51131.9
VanillaBackbone=LLaVA-1.5-13B, Vision token count=1442026.0610063.367.7—87.971.680——35.4———68.2—————63.81,53133.1
MS-ResamplerBackbone=LLaVA-1.5-7B, Vision token count=362026.0699.3261.264.9—87.269.976.8——32.8———62.8—————55.81,42932.6
SAEPBackbone=LLaVA-1.5-7B, Vision token count=1442026.0699.3161.863.8—87.26977.6——32.8———64.8—————56.41,42431.7
LDP-v2Backbone=LLaVA-1.5-7B, Vision token count=1442026.0699.2561.565.1—86.370.977.3——31.7———63.9—————56.41,43531.9
MS-ResamplerBackbone=LLaVA-1.5-13B, Vision token count=1442026.0698.7762.466.9—87.972.878.3——35.1———66.2—————58.51,54733.7
VisionZip 🔥Tokens Retained=192, Compression Ratio=↓ 66.7%2025.0898.7—————————————————————
MMTokTokens Retained=192, Compression Ratio=↓ 66.7%2025.0898.7—————————————————————
LDP-v2Backbone=LLaVA-1.5-7B, Vision token count=642026.0698.6160.166.2—86.871.476.1——30.1———62.6—————55.11,41433.5
ZOO-PruneRetained Tokens=192, Pruning Rate=66.7%2025.0998.659.9566.671,762.4186.7373.1278.6559.1165.56———37.33—————————
PRUNESID-DynEarly Cmp.=✓, Retained Tokens=192, Compression Ratio=66.7%2026.0398.6—————————————————————
PRUNESIDLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0398.5—————————————————————
TokenPackerBackbone=LLaVA-1.5-13B, Vision token count=1442026.0698.2262.367.4—87.470.778.8——34.5———65.4—————58.81,52533.9
ADSC# Vision Tokens=1922026.0298.261—1,8018569.577.156.2———1.8——————————
DivPruneTokens Retained=192, Compression Ratio=↓ 66.7%2025.0898.2—————————————————————
HoloVLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0398.2—————————————————————
AutoSelectLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0398.2—————————————————————
DivPruneRetained Tokens=192, Pruning Rate=66.7%2025.0998.159.4266.581,781.586.7672.8877.9858.4665.72———36.56—————————
PRUNESID-DynEarly Cmp.=✓, Retained Tokens=128, Compression Ratio=77.8%2026.0398.1—————————————————————
LDP-v2Backbone=LLaVA-1.5-13B, Vision token count=1442026.0698.036266.9—87.672.378.4——34.4———65.3—————58.41,53333.2
PRUNESIDBase Model=LLaVA-1.5 13B, Retained Tokens=192 (↓ 66.7%)2026.039859.665.91,77086.472.87858.6——56—36.965.2————————
VisionZipRetained Tokens=192, Pruning Rate=66.7%2025.0997.959.166.91,75485.173.578.159.565.2———36.4—————————
VisionZipTokens Retained=192, Compression Ratio=↓ 66.7%2025.0897.9—————————————————————
VisionZipLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0397.9—————————————————————
ZOO-PruneRetained Tokens=128, Pruning Rate=77.8%2025.0997.858.8967.011,791.186.9573.3877.8358.864.5———35.56—————————
DivPruneLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0397.8—————————————————————
VisionZipBase Model=LLaVA-1.5 13B, Retained Tokens=192 (↓ 66.7%)2026.0397.859.665.91,77086.472.87858.6——54.9—36.965.2————————
LDP-v2Backbone=LLaVA-1.5-7B, Vision token count=362026.0697.7759.964.8—86.469.775.2——31.7———61—————53.61,44332.3
AutoSelectLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0397.6—————————————————————
TokenPackerBackbone=LLaVA-1.5-7B, Vision token count=642026.0697.596164.4—86.569.477.1——29.1———63—————55.61,44031.3
FasterVLMBackbone=LLaVA-1.5-7B, Vision token count=1442026.0697.545864.3—83.567.876.2——32.2———62.3—————57.21,44031.3
MMTokTokens Retained=128, Compression Ratio=↓ 77.8%2025.0897.5—————————————————————
ADSC# Vision Tokens=1282026.0297.460.1—1,78284.569.976.355.1———2.6——————————
VisionZip 🔥Tokens Retained=128, Compression Ratio=↓ 77.8%2025.0897.4—————————————————————
PRUNESIDBase Model=LLaVA-1.5 13B, Retained Tokens=128 (↓ 77.8%)2026.0397.458.965.51,81185.973.176.757.5——56.8—35.864.1————————
TokenPackerBackbone=LLaVA-1.5-7B, Vision token count=362026.0697.3960.563.5—86.567.876.1——30.2———62.5—————55.21,42032.3
DARTLoc.=L2, Retained Tokens=192, Pruning Ratio=66.7%2026.0397.3—————————————————————
SAEPBackbone=LLaVA-1.5-7B, Vision token count=642026.0697.1660.362.9—866875.6——30.5———63.4—————53.91,38033.1
MQT-LLaVABackbone=LLaVA-1.5-7B, Vision token count=1442026.0697.0961.464.4—86.267.676.8——29.8———62—————52.61,44531.9
VisionZipRetained Tokens=128, Pruning Rate=77.8%2025.099757.966.71,74385.27476.858.763.8———36.1—————————
DivPruneRetained Tokens=128, Pruning Rate=77.8%2025.099758.8966.071,748.5686.5372.8377.158.1764.22———35.56—————————
VisionZipTokens Retained=128, Compression Ratio=↓ 77.8%2025.0897—————————————————————
VisionZipBase Model=LLaVA-1.5 13B, Retained Tokens=128 (↓ 77.8%)2026.039757.966.71,74385.27476.858.7——55—36.163.8————————
Pixel-ShuffleBackbone=LLaVA-1.5-7B, Vision token count=1442026.0696.9260.264.5—8769.675.6——29.1———61.6—————54.21,42231.7
DivPruneTokens Retained=128, Compression Ratio=↓ 77.8%2025.0896.9—————————————————————
DivPruneLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0396.9—————————————————————
PRUNESIDLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0396.9—————————————————————
HoloVLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0396.8—————————————————————
VisionZipLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0396.6—————————————————————
ZOO-PruneRetained Tokens=64, Pruning Rate=88.9%2025.0996.558.5864.781,780.0385.3472.0976.3958.5963.02———36—————————
AutoSelectLoc.=Pre, Retained Tokens=64, Pruning Ratio=88.9%2026.0396.5—————————————————————
MMTokTokens Retained=64, Compression Ratio=↓ 88.9%2025.0896.4—————————————————————
PRUNESID-DynEarly Cmp.=✓, Retained Tokens=64, Compression Ratio=88.9%2026.0396.3—————————————————————
PyramidDrop# Vision Tokens=1922026.0296.257.3—1,79782.369.275.156.5———3.8——————————
DARTLoc.=L2, Retained Tokens=128, Pruning Ratio=77.8%2026.0396.2—————————————————————
FasterVLMBackbone=LLaVA-1.5-13B, Vision token count=1442026.0696.1759.166.4—86.171.977.1——33.6———63.8—————58.81,48932.3
PDropLoc.=L8–L24, Retained Tokens=192, Pruning Ratio=66.7%2026.0396—————————————————————
Pixel-ShuffleBackbone=LLaVA-1.5-13B, Vision token count=1442026.0695.9661.266.2—86.971.376.9——33.1———64—————571,48332.1
C-AbstractorBackbone=LLaVA-1.5-7B, Vision token count=1442026.0695.9160.463.3—85.968.575.7——28.2———62.3—————54.11,39231.1
SparseVLMLoc.=L1–L32, Retained Tokens=192, Pruning Ratio=66.7%2026.0395.8—————————————————————
DivPruneRetained Tokens=64, Pruning Rate=88.9%2025.0995.457.6664.61,777.9384.871.3475.257.1162.44———35.22—————————
DivPruneTokens Retained=64, Compression Ratio=↓ 88.9%2025.0895.3—————————————————————
SAEPBackbone=LLaVA-1.5-7B, Vision token count=362026.0695.359.360.5—85.467.676.3——30.9———61.7—————51.71,35631.1
MQT-LLaVABackbone=LLaVA-1.5-7B, Vision token count=642026.0695.256063.5—83.66775.3——28.9———60.8—————51.61,45430.9
PyramidDrop# Vision Tokens=1282026.0295.257.1—1,76182.368.472.956.6———4.8——————————
PRUNESIDBase Model=LLaVA-1.5 13B, Retained Tokens=64 (↓ 88.9%)2026.0395.257.863.81,7118271.875.256.3——57.3—35.762.8————————
ADSC# Vision Tokens=642026.0295.158.4—1,7018469.873.953.2———4.9——————————
PRUNESIDLoc.=Pre, Retained Tokens=64, Pruning Ratio=88.9%2026.0395.1—————————————————————
DivPruneLoc.=Pre, Retained Tokens=64, Pruning Ratio=88.9%2026.0394.9—————————————————————
VisionZip 🔥Tokens Retained=64, Compression Ratio=↓ 88.9%2025.0894.8—————————————————————
HoloVLoc.=Pre, Retained Tokens=64, Pruning Ratio=88.9%2026.0394.8—————————————————————
MQT-LLaVABackbone=LLaVA-1.5-7B, Vision token count=362026.0694.736063.4—81.966.875.3——27.8———59—————511,46432.1
PDropLoc.=L8–L24, Retained Tokens=128, Pruning Ratio=77.8%2026.0394.7—————————————————————
VisionTrim ‡Token Count=16, Reduction Ratio=97.2%, Variant=Text-Guided Vision Complement2026.0194.35561.31,61278.170.272.355.455.930.250.5———————————
C-AbstractorBackbone=LLaVA-1.5-7B, Vision token count=642026.0694.2459.161.8—85.768.673.7——26.1———60—————51.31,41532.3
VisionZipBase Model=LLaVA-1.5 13B, Retained Tokens=64 (↓ 88.9%)2026.0394.256.264.91,6767674.473.757.4——55.9—36.460.4————————
Pixel-ShuffleBackbone=LLaVA-1.5-7B, Vision token count=642026.0694.1858.764.4—85.568.773.2——25.8———59.3—————52.21,38332.1
C-AbstractorBackbone=LLaVA-1.5-7B, Vision token count=362026.0693.9758.362.4—85.76972.4——29.3———58.4—————49.61,34731.8
VisionZipRetained Tokens=64, Pruning Rate=88.9%2025.0993.756.264.91,6767674.473.757.460.4———36.4—————————
VisionZipTokens Retained=64, Compression Ratio=↓ 88.9%2025.0893.7—————————————————————
HiREDLoc.=Pre, Retained Tokens=192, Pruning Ratio=66.7%2026.0393.7—————————————————————
FasterVLMBackbone=LLaVA-1.5-7B, Vision token count=642026.0693.485561.6—80.468.872.5——28.7———57.9—————561,34231.8
VisionZipLoc.=Pre, Retained Tokens=64, Pruning Ratio=88.9%2026.0392.7—————————————————————
SparseVLMLoc.=L1–L32, Retained Tokens=128, Pruning Ratio=77.8%2026.0392.4—————————————————————
DARTLoc.=L2, Retained Tokens=64, Pruning Ratio=88.9%2026.0392.3—————————————————————
HiREDLoc.=Pre, Retained Tokens=128, Pruning Ratio=77.8%2026.0392.2—————————————————————
VisionTrim ‡Token Count=8, Reduction Ratio=98.6%, Variant=Text-Guided Vision Complement2026.0191.851.758.31,52575.269.970.255.354.430.449.9———————————
C-AbstractorBackbone=LLaVA-1.5-13B, Vision token count=1442026.0691.7560.463—86.370.474.8——28.2———61—————53.91,39131.3
Pixel-ShuffleBackbone=LLaVA-1.5-7B, Vision token count=362026.0691.657.261.3—84.668.770.6——25.8———56.1—————48.11,30932.9
VisionTrimToken Count=16, Reduction Ratio=97.2%, Variant=Standard2026.019051.557.71,51072.669.468.45353.428.950———————————