Share your thoughts, 1 month free Claude Pro on usSee more

CommonSense Reasoning on Wino

77.4Accuracy

Full-precision

Updated 9d ago

Evaluation Results

Method	Links
Full-precision 2026.04		77.4
Full-precision 2026.04		76.4
PMQ 2026.04		75.3
Uniform 2026.04		75.14
Router norm + Max var 2026.04		74.98
PMQ 2026.04		74.98
Router norm + Max var 2026.04		74.66
PMQ 2026.04		74.66
PMQ 2026.04		74.66
Router norm + Max var 2026.04		74.35
PMQ 2026.04		74.19
Router norm + Max var 2026.04		73.95
Router norm + Max var 2026.04		73.8
PMQ 2026.04		73.24
Hessian 2026.04		72.97
PMQ 2026.04		72.3
Hessian 2026.04		71.25
PMQ 2026.04		71.19
Router norm + Max var 2026.04		70.17
Router norm + Max var 2026.04		69.93
Hessian 2026.04		69.11
SSQR-4% 2025.07		68.9
BF16 Baseline 2025.07		68.75
HPTQ 2025.07		68.35
SSQR-5% 2025.07		68.35
Router norm + Max var 2026.04		68.35
SSQR-3% 2025.07		68.27
SSQR-1% 2025.07		68.19
Uniform 2026.04		68.19
GPTQ 2025.07		68.11
HPTQ 2025.07		68.03
SSQR-2% 2025.07		68.03
SSQR-1% 2025.07		66.93
SSQR-4% 2025.07		66.77
SSQR-3% 2025.07		66.69
SSQR-5% 2025.07		66.69
GPTQ 2025.07		66.06
PCMind-2.1-Kaiyuan-2B 2025.12		65.75
SSQR-2% 2025.07		65.51
Qwen3-4B 2025.12		65.43
Router norm + Max var 2026.04		65.43
gemma2-2B 2025.12		65.35
PMQ 2026.04		65.35
Uniform 2026.04		65.27
SmolLM3-3B 2025.12		64.4
Router norm + Max var 2026.04		64.4
llama-3.2-3B 2025.12		64.09
Llama-3.1 2026.01		63.9
Qwen2.5-3B 2025.12		63.69
BSP 2026.04		62.19
PMQ 2026.04		61.96
YuLan-Mini-2.4B 2025.12		61.88
Router norm + Max var 2026.04		61.25
HPTQ 2025.07		60.85
OLMo-2-0425-1B 2025.12		60.38
A3 2026.01		60.2
Qwen2-1.5B 2025.12		59.83
PMQ 2026.04		59.83
Qwen2.5-1.5B 2025.12		59.67
Router norm + Max var 2026.04		59.59
Qwen3-1.7B 2025.12		59.27
6-LOOP 2025.12		59.19
SmolLM2-1.7B 2025.12		59.12
MOE 2025.12		58.8
SSQR-3% 2025.07		58.48
VERSATILEFFN 2025.12		58.17
llama-3.2-1B 2025.12		58.17
4-LOOP 2025.12		57.87
SSQR-4% 2025.07		57.85
Slim-LLM 2026.04		57.54
2-LOOP 2025.12		57.38
SSQR-5% 2025.07		57.38
PMQ 2026.04		57.38
RoPE 2025.12		56.7
DiffuLlama 2026.01		56.4
RoPE++EH 2025.12		55.9
VERSATILEFFN 2025.12		55.88
4-LOOP 2025.12		55.64
Qwen3-0.6B 2025.12		55.64
6-LOOP 2025.12		55.49
BASE 2025.12		55.49
2-LOOP 2025.12		55.41
RoPE++EC 2025.12		55.3
BASE 2025.12		55.01
A3 2026.01		54.3
MOE 2025.12		53.91
RoPE++EH 2025.12		53.6
VERSATILEFFN 2025.12		53.51
RoPE++EC 2025.12		53.3
6-LOOP 2025.12		53.2
RoPE 2025.12		53.2
SSQR-2% 2025.07		53.12
A3 2026.01		52.8
4-LOOP 2025.12		52.01
Dream 2026.01		51.8
2-LOOP 2025.12		51.78
SSQR-1% 2025.07		51.7
MOE 2025.12		51.38
Plaid 2026.01		51.3
BASE 2025.12		51.14

Showing 100 of 102 rows