List your company

Visual Question Answering on VQA v2 (test val)

78.5AccuracyNative
44.28453.16762.0570.933Mar 4, 2026Mar 21, 2026Apr 7, 2026Apr 24, 2026May 11, 2026May 28, 2026Jun 15, 2026

Evaluation results

MethodLinks
NativeRetained Tokens=5762026.0378.511176.5100———————
EvoPruneRetained Tokens=1282026.0376.81.10.770.8474.997.9———————
CDPrunerRetained Tokens=1282026.0376.61.410.770.9173.996.6———————
Firebolt-VLLLM=LFM2, Parameters=0.8B2026.0476.6————————————
MoE-LLaVALLM=Qwen, Parameters=2.2B2026.0476.2————————————
DivPruneRetained Tokens=1282026.03761.160.770.8572.995.3———————
VisPrunerRetained Tokens=1282026.0375.81.320.770.973.295.7———————
EvoPruneRetained Tokens=642026.0375.61.070.730.7772.995.3———————
FasterVLMRetained Tokens=1282026.0375.51.150.770.8572.394.6———————
CDPrunerRetained Tokens=642026.0375.41.380.730.8672.294.4———————
DARTRetained Tokens=1282026.0374.71.20.770.8870.992.7———————
DivPruneRetained Tokens=642026.0374.11.120.730.7970.492———————
VisPrunerRetained Tokens=642026.0373.91.350.730.8669.991.3———————
Mean NetworkBackbone=BEiT-3-base, Rel. Time=1.0×2026.0673.8—————7.6713.822.12.72—3.830.4
+ Temp. Sc.Backbone=BEiT-3-base, Rel. Time=1.0×2026.0673.8—————7.6713.822.12.72—3.740.4
StreamliningBackbone=BEiT-3-base, Rel. Time=2.0×2026.0673.8—————7.6713.822.12.72—3.790.4
CVPBackbone=BEiT-3-base, Rel. Time=2.3×2026.0673.8—————7.5717.324.62.7—3.170.4
CDPrunerRetained Tokens=322026.0373.61.220.710.869.691———————
4 samplesBackbone=BEiT-3-base, Rel. Time=4.0×2026.0673.5—————7.7615.622.72.71—2.150.4
2 samplesBackbone=BEiT-3-base, Rel. Time=2.0×2026.0673.3—————7.9413.721.82.74—2.530.4
EvoPruneRetained Tokens=322026.0373.11.060.710.7568.789.8———————
FasterVLMRetained Tokens=642026.0372.71.090.730.7968.289.2———————
VisPrunerRetained Tokens=322026.0372.11.350.710.8465.285.2———————
LLaVA-PhiLLM=Phi 2, Parameters=2.7B2026.0471.4————————————
DARTRetained Tokens=642026.0371.31.130.730.866.286.6———————
DivPruneRetained Tokens=322026.0371.21.10.710.7766.887.4———————
FasterVLMRetained Tokens=322026.0370.91.090.710.7764.584.4———————
CVPBackbone=ViLT, Rel. Time=2.2×2026.0669.3—————10.310.817.43.02—3.690.46
Mean NetworkBackbone=ViLT, Rel. Time=1.0×2026.0669.2—————10.532.413.33.1—7.720.48
+ Temp. Sc.Backbone=ViLT, Rel. Time=1.0×2026.0669.2—————10.532.513.33.1—7.680.48
StreamliningBackbone=ViLT, Rel. Time=2.0×2026.0669.2—————10.532.613.33.1—7.70.48
4 samplesBackbone=ViLT, Rel. Time=4.0×2026.0668.6—————10.8710.515.73.07—2.060.47
2 samplesBackbone=ViLT, Rel. Time=2.0×2026.0668—————11.437.613.23.13—2.90.48
DARTRetained Tokens=322026.0367.11.090.710.7659.878.1———————
OpenFlamingo v2LLM=MPT, Parameters=9.0B2026.0460.4————————————
IDEFICSLLM=LLaMA, Parameters=9.0B2026.0460————————————
FastVLLM=Vicuna, Parameters=7.0B2026.0455————————————
Kosmos-2LLM=GPT 2, Parameters=1.7B2026.0445.6————————————