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

Mathematical Reasoning on GSM8K (Exact Match %)

93.63Exact Match Accuracy (GSM8K)

Min-k

Updated 4d ago

Evaluation Results

Method	Links
Min-k 2026.04		93.63
Top-p 2026.04		93.56
Min-k 2026.04		93.56
Top-nσ 2026.04		93.4
Top-k 2026.04		93.33
Min-p 2026.04		93.18
Top-nσ 2026.04		93.18
Min-k 2026.04		93.18
Min-k 2026.04		93.1
Min-k 2026.04		93.1
Top-nσ 2026.04		93.03
Top-nσ 2026.04		92.87
Top-nσ 2026.04		92.87
Top-p 2026.04		92.8
Min-p 2026.04		92.8
Top-k 2026.04		92.72
Top-nσ 2026.04		92.72
Min-k 2026.04		92.65
Top-k 2026.04		92.49
Top-p 2026.04		91.74
Min-p 2026.04		91.74
Min-p 2026.04		89.92
Top-k 2026.04		89.61
Top-p 2026.04		88.76
Min-p 2026.04		85.14
Top-k 2026.04		79.08
Min-k 2026.04		77.39
Min-k 2026.04		76.65
Min-k 2026.04		76.48
Top-p 2026.04		76.27
Min-k 2026.04		76.15
Min-k 2026.04		76.02
Top-k 2026.04		75.44
Top-nσ 2026.04		75.44
Min-p 2026.04		75.36
Min-k 2026.04		74.79
Top-nσ 2026.04		74.07
Top-nσ 2026.04		73.77
Top-nσ 2026.04		73.69
Top-nσ 2026.04		73.09
Top-nσ 2026.04		72.55
Min-p 2026.04		62.7
Top-p 2026.04		56.25
Top-k 2026.04		54.13
Top-p 2026.04		46.63
Min-p 2026.04		28.81
Top-k 2026.04		9.86
Min-p 2026.04		3.94
Top-p 2026.04		2.5
Top-k 2026.04		1.06
Min-p 2026.04		0.45
Min-p 2026.04		0.38
Top-p 2026.04		0.15
Top-k 2026.04		0.08
Top-k 2026.04		0
Top-p 2026.04		0
Top-p 2026.04		0
Min-p 2026.04		0
Top-k 2026.04		0
Top-p 2026.04		0