List your company
ResearchFormerly WizWand

Zero-shot Evaluation on Various Tasks Average (test)

79.95Average AccuracyFP16
33.576445.615757.65569.6943Nov 27, 2025
Updated 8mo ago

Evaluation results

MethodLinks
FP16Model=LLaMA-3 70B, Precision=FP162025.1179.95—
FP16Model=LLaMA-2 70B, Precision=FP162025.1177.05—
SingleQuantModel=LLaMA-3 70B, Weight Quantization=RTN, Precision=4-bit2025.1176.3—
SingleQuantModel=LLaMA-2 70B, Weight Quantization=RTN, Precision=4-bit2025.1176.14—
SpinQuantModel=LLaMA-2 70B, Weight Quantization=GPTQ, Precision=4-bit2025.1176.06—
QuaRotModel=LLaMA-2 70B, Weight Quantization=GPTQ, Precision=4-bit2025.1175.68—
SpinQuantModel=LLaMA-2 70B, Weight Quantization=RTN, Precision=4-bit2025.1175.09—
QuaRotModel=LLaMA-2 70B, Weight Quantization=RTN, Precision=4-bit2025.1173.47—
FP16Model=LLaMA-3 8B, Precision=FP162025.1173.23—
DuQuantModel=LLaMA-3 70B, Weight Quantization=RTN, Precision=4-bit2025.1172.97—
FP16Model=LLaMA-2 13B, Precision=FP162025.1172.55—
SingleQuantModel=LLaMA-2 13B, Weight Quantization=RTN, Precision=4-bit2025.1171.5—
SpinQuantModel=LLaMA-3 70B, Weight Quantization=GPTQ, Precision=4-bit2025.1171.33—
SpinQuantModel=LLaMA-2 13B, Weight Quantization=GPTQ, Precision=4-bit2025.1170.93—
OSTQuantModel=LLaMA-2 13B, Weight Quantization=GPTQ, Precision=4-bit2025.1170.27—
FP16Model=LLaMA-2 7B, Precision=FP162025.1169.87—
DuQuantModel=LLaMA-2 70B, Weight Quantization=RTN, Precision=4-bit2025.1169.39—
OSTQuantModel=LLaMA-3 8B, Weight Quantization=GPTQ, Precision=4-bit2025.1169.08—
SingleQuantModel=LLaMA-3 8B, Weight Quantization=RTN, Precision=4-bit2025.1168.96—
QuaRotModel=LLaMA-2 13B, Weight Quantization=GPTQ, Precision=4-bit2025.1168.91—
SpinQuantModel=LLaMA-3 8B, Weight Quantization=GPTQ, Precision=4-bit2025.1168.7—
SpinQuantModel=LLaMA-2 13B, Weight Quantization=RTN, Precision=4-bit2025.1168.56—
SingleQuantModel=LLaMA-2 7B, Weight Quantization=RTN, Precision=4-bit2025.1167.18—
SpinQuantModel=LLaMA-3 8B, Weight Quantization=RTN, Precision=4-bit2025.1166.98—
OSTQuantModel=LLaMA-2 7B, Weight Quantization=GPTQ, Precision=4-bit2025.1166.39—
QuaRotModel=LLaMA-2 13B, Weight Quantization=RTN, Precision=4-bit2025.1166.25—
SpinQuantModel=LLaMA-2 7B, Weight Quantization=GPTQ, Precision=4-bit2025.1166.23—
QuaRotModel=LLaMA-3 8B, Weight Quantization=GPTQ, Precision=4-bit2025.1165.79—
DuQuantModel=LLaMA-3 8B, Weight Quantization=RTN, Precision=4-bit2025.1165.76—
SpinQuantModel=LLaMA-3 70B, Weight Quantization=RTN, Precision=4-bit2025.1165.66—
QuaRotModel=LLaMA-3 70B, Weight Quantization=GPTQ, Precision=4-bit2025.1165.37—
QuaRotModel=LLaMA-2 7B, Weight Quantization=GPTQ, Precision=4-bit2025.1165.01—
DuQuantModel=LLaMA-2 13B, Weight Quantization=RTN, Precision=4-bit2025.1164.98—
SpinQuantModel=LLaMA-2 7B, Weight Quantization=RTN, Precision=4-bit2025.1163.52—
QuaRotModel=LLaMA-3 8B, Weight Quantization=RTN, Precision=4-bit2025.1161.34—
DuQuantModel=LLaMA-2 7B, Weight Quantization=RTN, Precision=4-bit2025.1161.34—
QuaRotModel=LLaMA-2 7B, Weight Quantization=RTN, Precision=4-bit2025.1157.73—
QuaRotModel=LLaMA-3 70B, Weight Quantization=RTN, Precision=4-bit2025.1135.36—
AQLMBase Model=Llama-2-7B, #Bit=2.02, Evaluation Mode=Zero-shot2025.03—56.5
AQLMBase Model=Llama-2-13B, #Bit=1.97, Evaluation Mode=Zero-shot2025.03—60.6
BiLLMBase Model=Llama-3-70B, #Bit=1.1, Evaluation Mode=Zero-shot2025.03—44.2
ClusCompBase Model=Llama-2-7B, #Bit=2, Evaluation Mode=Zero-shot2025.03—56.6
ClusCompBase Model=Llama-2-13B, #Bit=1.99, Evaluation Mode=Zero-shot2025.03—62
ClusCompBase Model=Llama-3-8B, #Bit=2.01, Evaluation Mode=Zero-shot2025.03—53.6
ClusCompBase Model=Llama-3-70B, #Bit=1.14, Evaluation Mode=Zero-shot2025.03—39.7
ClusComp+Base Model=Llama-2-7B, #Bit=2, Evaluation Mode=Zero-shot2025.03—57.8
ClusComp+Base Model=Llama-2-13B, #Bit=1.99, Evaluation Mode=Zero-shot2025.03—63.1
ClusComp+Base Model=Llama-3-8B, #Bit=2.01, Evaluation Mode=Zero-shot2025.03—57.8
ClusComp+Base Model=Llama-3-70B, #Bit=1.14, Evaluation Mode=Zero-shot2025.03—51.4
DB-LLMBase Model=Llama-3-8B, #Bit=2, Evaluation Mode=Zero-shot2025.03—51.8
Llama-2-13BBase Model=Llama-2-13B, #Bit=16, Evaluation Mode=Zero-shot2025.03—67.8
Llama-2-7BBase Model=Llama-2-7B, #Bit=16, Evaluation Mode=Zero-shot2025.03—64.8
Llama-3-70BBase Model=Llama-3-70B, #Bit=16, Evaluation Mode=Zero-shot2025.03—75.4
Llama-3-8BBase Model=Llama-3-8B, #Bit=16, Evaluation Mode=Zero-shot2025.03—68.6
PB-LLMBase Model=Llama-3-8B, #Bit=2, Evaluation Mode=Zero-shot2025.03—38.8
PB-LLMBase Model=Llama-3-70B, #Bit=1.7, Evaluation Mode=Zero-shot2025.03—44.1
QuiPBase Model=Llama-3-8B, #Bit=2, Evaluation Mode=Zero-shot2025.03—36.8
QuIPBase Model=Llama-3-70B, #Bit=2, Evaluation Mode=Zero-shot2025.03—48.7
QuiP#Base Model=Llama-2-7B, #Bit=2.02, Evaluation Mode=Zero-shot2025.03—57.5
QuiP#Base Model=Llama-2-13B, #Bit=2.01, Evaluation Mode=Zero-shot2025.03—61.5