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

Commonsense Reasoning on CommonsenseQA (pass@1 accuracy)

86.6Accuracy (pass@1)

DENOISE

Updated 4d ago

Evaluation Results

Method	Links
DENOISE 2026.04		86.6
FACT-E 2026.04		86.2
BASE 2026.04		86
REFLECT 2026.04		86
SwiR 2025.10		85.34
POLISH 2026.04		84.8
FACT-E 2026.04		84
CoT 2025.10		83.95
CoT (Greedy) 2025.10		83.95
CONSIST. 2026.04		83.8
Soft Thinking 2025.10		83.7
BASE 2026.04		83
DENOISE 2026.04		82.8
FACT-E 2026.04		82.4
CONSIST. 2026.04		82
POLISH 2026.04		81.8
REFLECT 2026.04		81.8
CONSIST. 2026.04		81.8
BASE 2026.04		81.4
DENOISE 2026.04		81
POLISH 2026.04		81
REFLECT 2026.04		80.8
Zero-shot-CP 2024.03		73.81
Zero-shot 2024.03		71.33
Zero-shot-CoT 2024.03		68.96
Full Precision 2026.04		68.9
FACT-E 2026.04		63.4
CONSIST. 2026.04		55.6
REFLECT 2026.04		53.2
POLISH 2026.04		52.22
DENOISE 2026.04		51.6
BASE 2026.04		49.6
Bitnet 2026.04		47.2
BWTA 2026.04		44.8
Bitnet 2026.04		44.4
BWTA 2026.04		43.4
BWN 2026.04		42.5
BiLLM 2026.04		42.3
Gen-SSD 2026.04		42.07
Standard KD 2026.04		41.66
MCC-KD 2026.04		40.67
MoRSD 2026.04		40.43
BWN 2026.04		40.4
PB-LLM 2026.04		40
GPTQ 2026.04		37.5