Commonsense Reasoning
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
88.5BoolQ Accuracy
10
Feb 26, 2026
66.14Exact Match (EM)
10
Mar 18, 2026
95.7Accuracy
10
Feb 26, 2026
70.16HellaSwag Accuracy
9
May 13, 2026
79.2FRS
9
Apr 15, 2026
0.8Accuracy Delta
9
Apr 7, 2026
4.7Accuracy Delta
9
Apr 7, 2026
82.52HSwag Accuracy
9
Apr 1, 2026
84.98PIQA
9
Feb 26, 2026
81.6ACC
9
Feb 26, 2026
73.58Accuracy (5%)
9
Feb 26, 2026
71.52Accuracy (5%)
9
Feb 26, 2026
85.5Accuracy
9
Feb 26, 2026
64.2Accuracy
9
Feb 27, 2026
83.9Accuracy (SciQ zero-shot)
8
Jun 23, 2026
67Accuracy
8
Jun 15, 2026
85.5Pass@1 Accuracy
8
Jun 8, 2026
3.23ECE
8
Jun 5, 2026
78.5Score
8
Jul 7, 2026
50.95Accuracy
8
May 14, 2026
63.1Accuracy
8
May 13, 2026
60Accuracy
8
Apr 28, 2026
48.88Accuracy (SWAG Generalization)
8
Apr 1, 2026
62.4BoolQ Accuracy
8
Mar 31, 2026
52.58Accuracy
8
Mar 6, 2026