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

Text Generation Accuracy on GSM8K

86.4Accuracy

LoRA + PCGrad

10.417630.143849.8769.5962May 28, 2026Jun 1, 2026Jun 5, 2026Jun 9, 2026Jun 13, 2026Jun 17, 2026Jun 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
86.4
2026.06
85.53
2026.06
85.45
2026.06
85.45
2026.05
84.99
2026.05
84.99
2026.06
84.46
2026.06
82.94
2026.06
82.18
2026.05
81.8
2026.05
81.8
2026.05
81.43
2026.05
81.35
2026.05
81.35
2026.05
80.52
2026.05
80.52
2026.06
79.85
2026.05
79.76
2026.05
79.76
2026.06
79.63
2026.05
79.45
2026.06
79.24
2026.06
79.12
2026.05
78.85
2026.05
78.85
2026.06
78.77
2026.06
78.77
2026.05
78.17
2026.05
78.17
2026.06
76.42
2026.06
74.77
2026.06
74.63
2026.05
74.45
2026.05
74.45
2026.06
74.43
2026.05
74.22
2026.06
73.98
2026.05
73.92
2026.05
73.92
2026.06
73.9
2026.05
73.84
2026.05
73.84
2026.06
72.88
2026.06
72.87
2026.05
55.42
2026.05
55.02
2026.05
54.89
2026.05
54.74
2026.05
54.36
2026.05
53.39
2026.06
49.67
2026.06
48.65
2026.06
48.2
2026.06
48.14
2026.06
47.95
2026.06
47.53
2026.06
46.49
2026.05
15.04
2026.05
14.86
2026.05
14.74
2026.05
14.7
2026.05
13.76
2026.05
13.34