List your company
ResearchFormerly WizWand

Language Modeling on WikiText-2 raw (test)

5.674PerplexityBase
-1,714.099049,894.3689821,502.83733,111.30502Jan 2, 2023Jul 28, 2023Feb 21, 2024Sep 16, 2024Apr 11, 2025Nov 5, 2025Jun 1, 2026
Updated 5mo ago

Evaluation results

MethodLinks
BaseModel=Qwen3-VL-8B-Instruct2026.015.67400
SRAModel=Qwen3-VL-8B-Instruct2026.015.6780.0040.016
StandardModel=Qwen3-VL-8B-Instruct2026.016.3320.6581.337
SRAModel=Qwen3-VL-4B-Instruct2026.016.382-0.0240.044
BaseModel=Qwen3-VL-4B-Instruct2026.016.40600
StandardModel=Qwen3-VL-4B-Instruct2026.016.8370.4312.088
BaseModel=Qwen3-VL-2B-Instruct2026.017.26600
SRAModel=Qwen3-VL-2B-Instruct2026.017.2940.0280.018
SparseGPTModel size=175B, Sparsity=50%2023.018.21——
DenseModel size=175B, Sparsity=0%2023.018.35——
SparseGPTModel size=175B, Sparsity=4:82023.018.45——
SparseGPTModel size=175B, Sparsity=2:42023.018.74——
StandardModel=Qwen3-VL-2B-Instruct2026.018.8341.5680.622
SparseGPTModel size=66B, Sparsity=50%2023.019.32——
DenseModel size=66B, Sparsity=0%2023.019.34——
DenseModel size=30B, Sparsity=0%2023.019.56——
SparseGPTModel size=66B, Sparsity=4:82023.019.65——
SparseGPTModel size=30B, Sparsity=50%2023.019.79——
SparseGPTModel size=66B, Sparsity=2:42023.0110.09——
DenseModel size=13B, Sparsity=0%2023.0110.13——
SparseGPTModel size=30B, Sparsity=4:82023.0110.3——
DenseModel size=6.7B, Sparsity=0%2023.0110.86——
SparseGPTModel size=30B, Sparsity=2:42023.0110.9——
SUBFITSparsity=20%, Backbone=Llama-3.1-8B2026.0610.9——
SparseGPTModel size=13B, Sparsity=50%2023.0111.17——
SparseGPTModel size=6.7B, Sparsity=50%2023.0111.55——
SparseGPTModel size=13B, Sparsity=4:82023.0111.77——
DenseModel size=2.7B, Sparsity=0%2023.0112.47——
SparseGPTModel size=6.7B, Sparsity=4:82023.0112.56——
SparseGPTModel size=13B, Sparsity=2:42023.0112.96——
SparseGPTModel size=2.7B, Sparsity=50%2023.0113.48——
SUBFITSparsity=20%, Backbone=Llama-3.2-3B2026.0613.84——
BaseModel=Ministral-14B-Instruct-25122026.0114.1800
SparseGPTModel size=6.7B, Sparsity=2:42023.0114.2——
SRAModel=Ministral-14B-Instruct-25122026.0114.220.040.026
DenseModel size=1.3B, Sparsity=0%2023.0114.62——
SparseGPTModel size=2.7B, Sparsity=4:82023.0114.98——
StandardModel=Ministral-14B-Instruct-25122026.0115.1810.723
SparseGPTModel size=2.7B, Sparsity=2:42023.0117.18——
SparseGPTModel size=1.3B, Sparsity=50%2023.0117.46——
BaseModel=Ministral-3B-Instruct-25122026.0120.7300
StandardModel=Ministral-3B-Instruct-25122026.0120.780.050.097
SRAModel=Ministral-3B-Instruct-25122026.0120.790.060.018
DenseModel size=350M, Sparsity=0%2023.0122——
SUBFITSparsity=30%, Backbone=Llama-3.1-8B2026.0623.4——
SUBFITSparsity=30%, Backbone=Llama-3.2-3B2026.0624.35——
DenseModel size=125M, Sparsity=0%2023.0127.66——
SparseGPTModel size=350M, Sparsity=50%2023.0131.58——
AdaPruneModel size=1.3B, Sparsity=50%2023.0132.52——
SparseGPTModel size=125M, Sparsity=50%2023.0136.85——
AdaPruneModel size=350M, Sparsity=50%2023.0148.46——
SUBFITSparsity=37.5%, Backbone=Llama-3.2-3B2026.0658.62——
AdaPruneModel size=125M, Sparsity=50%2023.0158.66——
SUBFITSparsity=37.5%, Backbone=Llama-3.1-8B2026.0662.07——
MagnitudeModel size=350M, Sparsity=50%2023.0197.8——
MagnitudeModel size=30B, Sparsity=50%2023.01168——
MagnitudeModel size=125M, Sparsity=50%2023.01193——
MagnitudeModel size=2.7B, Sparsity=50%2023.01265——
MagnitudeModel size=6.7B, Sparsity=50%2023.01969——
MagnitudeModel size=66B, Sparsity=50%2023.014,200——
MagnitudeModel size=13B, Sparsity=50%2023.0112,000——
MagnitudeModel size=1.3B, Sparsity=50%2023.0117,000——
MagnitudeModel size=175B, Sparsity=50%2023.0143,000——