Physical Commonsense Reasoning
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
94.9Accuracy
724
Jun 24, 2026
83Accuracy
118
May 19, 2026
81.5Accuracy (PIQA)
99
May 29, 2026
85.91Accuracy
85
Jun 12, 2026
90.7Accuracy
71
Jun 4, 2026
7,497Accuracy
56
Mar 4, 2026
82.54Accuracy
51
Jun 29, 2026
76.56Accuracy
45
Apr 14, 2026
81.08Accuracy (%)
36
Jun 18, 2026
0Delta Accuracy
24
Feb 26, 2026
81.1PIQA Score
16
May 29, 2026
0.152Mean Per-Step Regret
15
Feb 26, 2026
81.3ACC
14
Jul 8, 2026
93.74Accuracy
14
Jul 9, 2026
75.52Character-level Accuracy
11
Mar 18, 2026
79.8Accuracy (PIQA)
10
Jun 1, 2026
68Accuracy
10
Mar 4, 2026
81.8Accuracy
8
May 13, 2026
84Accuracy
8
Apr 28, 2026
79.42Accuracy
8
May 21, 2026
79.8Accuracy (0-shot)
6
Feb 26, 2026
86.4Score
4
Mar 6, 2026
77.8Accuracy
3
May 25, 2026
1.89Average Relative Improvement
3
Feb 26, 2026
0.806PIQA Score
2
May 15, 2026