List your company
ResearchFormerly WizWand

Language Model Calibration on WikiText-2

5.82Perplexity (PPL)RIA+SQ+EBFT
5.28768.881312.47516.0687Jul 3, 2025
Updated 6mo ago

Evaluation results

MethodLinks
RIA+SQ+EBFTBase Model=Mistral, Outliers=16:256, Sparsity=8:162025.075.82——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=8:256, Sparsity=8:162025.075.96——————
RIA+SQBase Model=Mistral, Outliers=16:256, Sparsity=8:162025.076.06——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=4:256, Sparsity=8:162025.076.06——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=None, Sparsity=8:162025.076.14——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=16:256, Sparsity=2:42025.076.27——————
RIA+SQBase Model=Mistral, Outliers=8:256, Sparsity=8:162025.076.34——————
RIA+SQBase Model=Mistral, Outliers=4:256, Sparsity=8:162025.076.52——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=8:256, Sparsity=2:42025.076.59——————
RIA+SQBase Model=Mistral, Outliers=None, Sparsity=8:162025.076.74——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=4:256, Sparsity=2:42025.076.86——————
RIA+SQBase Model=Mistral, Outliers=16:256, Sparsity=2:42025.076.97——————
RIA+SQ+EBFTBase Model=Mistral, Outliers=None, Sparsity=2:42025.077.1——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=8:162025.077.4——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=8:162025.077.69——————
RIA+SQBase Model=Mistral, Outliers=8:256, Sparsity=2:42025.077.8——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=8:162025.077.88——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=None, Sparsity=8:162025.078.08——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=2:42025.078.24——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=8:162025.078.42——————
RIA+SQBase Model=Mistral, Outliers=4:256, Sparsity=2:42025.078.48——————
RIA+SQBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=8:162025.078.52——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=8:162025.079.11——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=2:42025.079.18——————
RIA+SQBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=8:162025.079.32——————
RIA+SQBase Model=Mistral, Outliers=None, Sparsity=2:42025.079.34——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=8:162025.079.52——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=2:42025.079.53——————
RIA+SQBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=8:162025.079.85——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=None, Sparsity=8:162025.0710.09——————
RIA+SQ+VC+EBFTBase Model=LLaMA3-8B, Outliers=None, Sparsity=2:42025.0710.49——————
RIA+SQBase Model=LLaMA3-8B, Outliers=None, Sparsity=8:162025.0710.55——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=2:42025.0710.68——————
RIA+SQBase Model=LLaMA3-8B, Outliers=16:256, Sparsity=2:42025.0711.48——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=2:42025.0712.73——————
RIA+SQBase Model=LLaMA3-8B, Outliers=8:256, Sparsity=2:42025.0714.28——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=2:42025.0714.47——————
RIA+SQ+VCBase Model=LLaMA3-8B, Outliers=None, Sparsity=2:42025.0716.19——————
RIA+SQBase Model=LLaMA3-8B, Outliers=4:256, Sparsity=2:42025.0716.5——————
RIA+SQBase Model=LLaMA3-8B, Outliers=None, Sparsity=2:42025.0719.13——————
GPT-2Model=GPT-2+CALM2026.04—6.21.186.72.810.83.87