List your company

CommonSense Reasoning on Wino

86.55AccuracyFFT
55.807663.788871.7779.7512Jul 24, 2025Sep 4, 2025Oct 17, 2025Nov 29, 2025Jan 11, 2026Feb 23, 2026Apr 7, 2026

Evaluation results

MethodLinks
FFTModel=Llama-3.2-3B, #Params=3.21B2025.0986.55
ABBAModel=Llama-3.2-3B, #Params=48.63M2025.0986.53
BoHAModel=Llama-3.2-3B, #Params=48.63M2025.0986.4
HiRAModel=Llama-3.2-3B, #Params=48.63M2025.0985.71
GraLoRAModel=Llama-3.2-3B, #Params=48.63M2025.0984.98
DoRAModel=Llama-3.2-3B, #Params=49.40M2025.0984.14
LoRAModel=Llama-3.2-3B, #Params=48.63M2025.0983.5
PiSSAModel=Llama-3.2-3B, #Params=48.63M2025.0983.24
rsLoRAModel=Llama-3.2-3B, #Params=48.63M2025.0982.02
Full-precisionModel=Mixtral 8x22B, Avg. bits/exp.=16 (FP), Memory (GB)=281.2, Evaluation protocol=zero-shot2026.0477.4
Full-precisionModel=Mixtral 8x7B, Avg. bits/exp.=16 (FP), Memory (GB)=96.8, Evaluation protocol=zero-shot2026.0476.4
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.75, Memory (GB)=17.7, Evaluation protocol=zero-shot2026.0475.3
UniformModel=Mixtral 8x7B, Avg. bits/exp.=3, Memory (GB)=19.3, Evaluation protocol=zero-shot2026.0475.14
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.75, Memory (GB)=17.7, Evaluation protocol=zero-shot2026.0474.98
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.5, Memory (GB)=16.1, Evaluation protocol=zero-shot2026.0474.98
FFTModel=Llama-3.2-1B, #Params=1.24B2025.0974.79
ABBAModel=Llama-3.2-1B, #Params=22.54M2025.0974.74
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.625, Memory (GB)=16.9, Evaluation protocol=zero-shot2026.0474.66
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.625, Memory (GB)=16.9, Evaluation protocol=zero-shot2026.0474.66
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.375, Memory (GB)=15.3, Evaluation protocol=zero-shot2026.0474.66
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.5, Memory (GB)=16.1, Evaluation protocol=zero-shot2026.0474.35
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.25, Memory (GB)=14.5, Evaluation protocol=zero-shot2026.0474.19
PiSSAModel=Llama-3.2-1B, #Params=22.54M2025.0974.03
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.25, Memory (GB)=14.5, Evaluation protocol=zero-shot2026.0473.95
GraLoRAModel=Llama-3.2-1B, #Params=22.54M2025.0973.95
rsLoRAModel=Llama-3.2-1B, #Params=22.54M2025.0973.85
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.375, Memory (GB)=15.3, Evaluation protocol=zero-shot2026.0473.8
BoHAModel=Llama-3.2-1B, #Params=22.54M2025.0973.59
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.0, Memory (GB)=13.1, Evaluation protocol=zero-shot2026.0473.24
DoRAModel=Llama-3.2-1B, #Params=22.92M2025.0973.01
HessianModel=Mixtral 8x7B, Avg. bits/exp.=2.5, Memory (GB)=17.0, Evaluation protocol=zero-shot2026.0472.97
LoRAModel=Llama-3.2-1B, #Params=22.54M2025.0972.38
PMQModel=Mixtral 8x7B, Avg. bits/exp.=2.125, Memory (GB)=13.8, Evaluation protocol=zero-shot2026.0472.3
HiRAModel=Llama-3.2-1B, #Params=22.54M2025.0971.27
HessianModel=Mixtral 8x7B, Avg. bits/exp.=2.25, Memory (GB)=15.3, Evaluation protocol=zero-shot2026.0471.25
PMQModel=Mixtral 8x7B, Avg. bits/exp.=1.75, Memory (GB)=11.7, Evaluation protocol=zero-shot2026.0471.19
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.125, Memory (GB)=13.8, Evaluation protocol=zero-shot2026.0470.17
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=2.0, Memory (GB)=13.1, Evaluation protocol=zero-shot2026.0469.93
HessianModel=Mixtral 8x7B, Avg. bits/exp.=2.0, Memory (GB)=13.6, Evaluation protocol=zero-shot2026.0469.11
SSQR-4%Avg. Bitwidth=5.405, Zero-shot=true2025.0768.9
BF16 BaselineAvg. Bitwidth=16.000, Zero-shot=true2025.0768.75
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=12 + 22025.0968.51
HPTQAvg. Bitwidth=3.125, Zero-shot=true2025.0768.35
SSQR-5%Avg. Bitwidth=5.725, Zero-shot=true2025.0768.35
Router norm + Max varModel=Mixtral 8x7B, Avg. bits/exp.=1.75, Memory (GB)=11.7, Evaluation protocol=zero-shot2026.0468.35
SSQR-3%Avg. Bitwidth=5.085, Zero-shot=true2025.0768.27
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=6 + 22025.0968.27
SSQR-1%Avg. Bitwidth=4.445, Zero-shot=true2025.0768.19
UniformModel=Mixtral 8x22B, Avg. bits/exp.=3, Memory (GB)=57.5, Evaluation protocol=zero-shot2026.0468.19
GPTQAvg. Bitwidth=4.125, Zero-shot=true2025.0768.11
HPTQAvg. Bitwidth=4.125, Zero-shot=true2025.0768.03
SSQR-2%Avg. Bitwidth=4.765, Zero-shot=true2025.0768.03
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=12 + 22025.0967.09
SSQR-1%Avg. Bitwidth=3.445, Zero-shot=true2025.0766.93
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=6 + 22025.0966.93
SSQR-4%Avg. Bitwidth=4.405, Zero-shot=true2025.0766.77
SSQR-3%Avg. Bitwidth=4.085, Zero-shot=true2025.0766.69
SSQR-5%Avg. Bitwidth=4.725, Zero-shot=true2025.0766.69
GPTQAvg. Bitwidth=3.125, Zero-shot=true2025.0766.06
PCMind-2.1-Kaiyuan-2BParams=2B2025.1265.75
SSQR-2%Avg. Bitwidth=3.765, Zero-shot=true2025.0765.51
Qwen3-4BParams=4B2025.1265.43
Router norm + Max varModel=Mixtral 8x22B, Avg. bits/exp.=2.25, Memory (GB)=43.0, Evaluation protocol=zero-shot2026.0465.43
gemma2-2BParams=2B2025.1265.35
PMQModel=Mixtral 8x22B, Avg. bits/exp.=2.5, Memory (GB)=46.7, Evaluation protocol=zero-shot2026.0465.35
EMoEBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=3 + 22025.0965.35
UniformModel=Mixtral 8x7B, Avg. bits/exp.=2, Memory (GB)=13.1, Evaluation protocol=zero-shot2026.0465.27
SmolLM3-3BParams=3B2025.1264.4
Router norm + Max varModel=Mixtral 8x22B, Avg. bits/exp.=2.5, Memory (GB)=46.7, Evaluation protocol=zero-shot2026.0464.4
Top-kBase Model=ERNIE-4.5-21B-A3B, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=3 + 22025.0964.17
llama-3.2-3BParams=3B2025.1264.09
Llama-3.1Size=8B, Type=AR2026.0163.9
Qwen2.5-3BParams=3B2025.1263.69
BSPModel=Mixtral 8x7B, Avg. bits/exp.=2.5, Memory (GB)=17.0, Evaluation protocol=zero-shot2026.0462.19
PMQModel=Mixtral 8x22B, Avg. bits/exp.=2.25, Memory (GB)=43.0, Evaluation protocol=zero-shot2026.0461.96
YuLan-Mini-2.4BParams=2.4B2025.1261.88
Router norm + Max varModel=Mixtral 8x22B, Avg. bits/exp.=2.0, Memory (GB)=38.6, Evaluation protocol=zero-shot2026.0461.25
HPTQAvg. Bitwidth=2.125, Zero-shot=true2025.0760.85
OLMo-2-0425-1BParams=1B2025.1260.38
A3Size=8B, Type=A32026.0160.2
Qwen2-1.5BParams=1.5B2025.1259.83
PMQModel=Mixtral 8x22B, Avg. bits/exp.=2.0, Memory (GB)=38.6, Evaluation protocol=zero-shot2026.0459.83
Qwen2.5-1.5BParams=1.5B2025.1259.67
Router norm + Max varModel=Mixtral 8x22B, Avg. bits/exp.=1.75, Memory (GB)=35.2, Evaluation protocol=zero-shot2026.0459.59
Qwen3-1.7BParams=1.7B2025.1259.27
6-LOOPBase Model Params=1.21B, Protocol=Zero-shot2025.1259.19
SmolLM2-1.7BParams=1.7B2025.1259.12
MOEBase Model Params=1.21B, Protocol=Zero-shot2025.1258.8
SSQR-3%Avg. Bitwidth=3.085, Zero-shot=true2025.0758.48
VERSATILEFFNBase Model Params=1.21B, Protocol=Zero-shot2025.1258.17
llama-3.2-1BParams=1B2025.1258.17
4-LOOPBase Model Params=1.21B, Protocol=Zero-shot2025.1257.87
SSQR-4%Avg. Bitwidth=3.405, Zero-shot=true2025.0757.85
Slim-LLMModel=Mixtral 8x7B, Avg. bits/exp.=2.0, Memory (GB)=13.6, Evaluation protocol=zero-shot2026.0457.54
2-LOOPBase Model Params=1.21B, Protocol=Zero-shot2025.1257.38
SSQR-5%Avg. Bitwidth=3.725, Zero-shot=true2025.0757.38
PMQModel=Mixtral 8x22B, Avg. bits/exp.=1.75, Memory (GB)=35.2, Evaluation protocol=zero-shot2026.0457.38
Top-kBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=6 + 22025.0957.22
Top-kBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=Top-k, Inference Budget (k')=12 + 22025.0957.22
EMoEBase Model=DeepSeek-V2-Lite, Trained Activated Experts=6+2, MoE Routing Method=EMoE, Inference Budget (k')=12 + 22025.0956.99