List your company
ResearchFormerly WizWand

Commonsense Reasoning on Winogrande (HS, Acc)

82.35AccuracyNone
53.261260.813168.36575.9169Nov 26, 2025Dec 31, 2025Feb 4, 2026Mar 11, 2026Apr 15, 2026May 20, 2026Jun 25, 2026
Updated 4mo ago

Evaluation results

MethodLinks
NoneModel=Mistral, Mixing ratio (r)=0.12026.0382.3547.75
NoneModel=Llama3, Mixing ratio (r)=0.12026.0382.225.25
BeavertailsModel=Llama3, Mixing ratio (r)=0.12026.0381.9829.18
GR-SAPModel=Llama3, Mixing ratio (r)=0.12026.0381.640.68
AegisModel=Llama3, Mixing ratio (r)=0.12026.0381.611.42
AegisModel=Mistral, Mixing ratio (r)=0.12026.0380.779
NoneModel=Qwen2.5, Mixing ratio (r)=0.12026.0379.3217.3
AegisModel=Qwen2.5, Mixing ratio (r)=0.12026.0377.9314.42
BeavertailsModel=Qwen2.5, Mixing ratio (r)=0.12026.0377.9340.5
GR-SAPModel=Qwen2.5, Mixing ratio (r)=0.12026.0377.5311.04
NoneModel=OLMo2, Mixing ratio (r)=0.12026.0376.060.79
GR-SAPModel=OLMo2, Mixing ratio (r)=0.12026.0375.190.45
BeavertailsModel=OLMo2, Mixing ratio (r)=0.12026.0375.0921.99
OriginalModel=OLMo2, Mixing ratio (r)=0.12026.0375.060.5
AegisModel=OLMo2, Mixing ratio (r)=0.12026.0374.982.15
GRINQH-6b-RTNEff. Bits=3.881, Model=Llama-3.1-8B-Instruct2026.0674.66—
GRINQH-8b-RTNEff. Bits=3.96, Model=Llama-3.1-8B-Instruct2026.0674.35—
MatGPTQ-EP-Mix’n’MatchEff. Bits=4.00, Model=Llama-3.1-8B-Instruct2026.0673.56—
BF16-BaselineEff. Bits=16, Model=Llama-3.1-8B-Instruct2026.0673.48—
GRINQH-6b-RTNEff. Bits=2.98, Model=Llama-3.1-8B-Instruct2026.0673.16—
GR-SAPModel=Mistral, Mixing ratio (r)=0.12026.0373.0912.33
GRINQH-8b-RTNEff. Bits=2.95, Model=Llama-3.1-8B-Instruct2026.0673.01—
MatGPTQ-EP-Mix’n’MatchEff. Bits=3.00, Model=Llama-3.1-8B-Instruct2026.0672.53—
GRINQH-4b-GPTQEff. Bits=3.01, Model=Llama-3.1-8B-Instruct2026.0672.53—
MatGPTQEff. Bits=4.00, Model=Llama-3.1-8B-Instruct2026.0672.06—
MatGPTQEff. Bits=3.00, Model=Llama-3.1-8B-Instruct2026.0671.27—
BeavertailsModel=Mistral, Mixing ratio (r)=0.12026.0369.9841.98
GRINQH-8b-RTNEff. Bits=2.24, Model=Qwen3-8B2026.0667.72—
GRINQH-6b-RTNEff. Bits=3.05, Model=Qwen3-8B2026.0667.64—
BF16-BaselineEff. Bits=16, Model=Qwen3-8B2026.0667.56—
GRINQH-8b-RTNEff. Bits=2.08, Model=Llama-3.1-8B-Instruct2026.0667.4—
GRINQH-4b-GPTQEff. Bits=2.09, Model=Qwen3-8B2026.0666.93—
GRINQH-4b-GPTQEff. Bits=3.02, Model=Qwen3-8B2026.0666.61—
GRINQH-8b-RTNEff. Bits=3.06, Model=Qwen3-8B2026.0665.9—
MatGPTQEff. Bits=3.00, Model=Qwen3-8B2026.0664.64—
MatGPTQ-EP-Mix’n’MatchEff. Bits=3.00, Model=Qwen3-8B2026.0664.48—
Gated KalmaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1164.17—
Gated DeltaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1162.35—
Mamba2evaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1162.19—
Transformerevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1161.72—
DeltaNetevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1161.72—
CARVE + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0660.71—
Muon + PCZero-shot=true, Model=Llama-1B2026.0659.67—
GDN-2 + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0658.56—
CARVEModel Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0658.12—
Muon baselineZero-shot=true, Model=Llama-1B2026.0658.09—
GDN-2Model Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0657.85—
KDA + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0657.77—
Mamba-3 SISO + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0657.3—
Mamba-3 MIMO + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0657.06—
Gated DeltaNet + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0656.83—
Gated DeltaNetModel Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0656.75—
AdamW + PCZero-shot=true, Model=Llama-1B2026.0656.27—
Mamba-3 SISOModel Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0656.2—
Mamba-2 + SWAModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0656.17—
TransformerModel Category=Hybrid, Parameter Scale=1.3B, Token Scale=100B2026.0655.85—
Mamba-3 MIMOModel Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0655.78—
KDAModel Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0655.72—
Mamba-2Model Category=Recurrent, Parameter Scale=1.3B, Token Scale=100B2026.0655.33—
Gated Linear Attentionevaluation_mode=zero-shot, parameters=2.8B, context_length=< 2K tokens2025.1154.54—
AdamW baselineZero-shot=true, Model=Llama-1B2026.0654.38—