List your company

Language Modeling on C4 Llama-3.1-8B-Instruct (val)

9.09PerplexityBF16 Baseline
6.600423.405240.2157.0148Jul 24, 2025

Evaluation results

MethodLinks
BF16 BaselineAvg. Bitwidth=16.0002025.079.09
HPTQAvg. Bitwidth=4.1252025.079.99
SSQR-5%Avg. Bitwidth=5.7252025.0710.03
SSQR-4%Avg. Bitwidth=5.4052025.0710.05
SSQR-3%Avg. Bitwidth=5.0852025.0710.11
SSQR-2%Avg. Bitwidth=4.7652025.0710.2
SSQR-1%Avg. Bitwidth=4.4452025.0710.3
GPTQAvg. Bitwidth=4.1252025.0710.46
HPTQAvg. Bitwidth=3.1252025.0711.04
SSQR-5%Avg. Bitwidth=4.7252025.0711.13
SSQR-4%Avg. Bitwidth=4.4052025.0711.31
SSQR-3%Avg. Bitwidth=4.0852025.0711.54
SSQR-2%Avg. Bitwidth=3.7652025.0711.75
SSQR-1%Avg. Bitwidth=3.4452025.0712.35
GPTQAvg. Bitwidth=3.1252025.0713.16
SSQR-5%Avg. Bitwidth=3.7252025.0716.13
HPTQAvg. Bitwidth=2.1252025.0716.37
SSQR-4%Avg. Bitwidth=3.4052025.0717.77
SSQR-3%Avg. Bitwidth=3.0852025.0720.03
SSQR-2%Avg. Bitwidth=2.7652025.0725.21
SSQR-1%Avg. Bitwidth=2.4452025.0739.96
GPTQAvg. Bitwidth=2.1252025.0771.33