Numerical Reasoning
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
70.31Exact Match (EM)
66
Apr 15, 2026
74.5Hit Rate @ 1
24
Apr 21, 2026
63.6Hit Rate @1
24
Apr 21, 2026
43.9H@1
24
Apr 21, 2026
92.9Accuracy
14
Mar 10, 2026
64.48Accuracy
13
May 21, 2026
98.9CD4
13
Mar 18, 2026
85.41Precision
8
Mar 10, 2026
78.65Precision
8
Mar 10, 2026
70Accuracy (Error <= 1)
6
May 21, 2026
1.13MAE (Scale 1)
6
May 21, 2026
85.2EM
6
Mar 6, 2026
83.74EM
4
Mar 6, 2026
57.4Exact Match (EM)
4
Feb 27, 2026
72.4Exact Match
4
Feb 26, 2026
63.4Accuracy
4
Feb 27, 2026
74.2Accuracy
4
Feb 27, 2026
67.5Exact Match
4
Feb 27, 2026
59.1Exact Match (EM)
4
Feb 27, 2026
68.7EM
4
Feb 27, 2026
79.8EM
4
Feb 27, 2026