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

Reasoning on ARC-E, CSQA, Lambada, PIQA, Wino

80.85ARC-E Accuracy

Base

44.91854.246563.57572.9035Oct 7, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.10
80.8556.2769.4980.6374.1972.29
2025.10
79.4646.670.3379.1172.369.56
2025.10
76.333.0168.2578.0769.0664.94
2025.10
73.3625.1460.7873.2369.3860.38
2025.10
72.3121.0552.873.0766.4657.14
2025.10
7125.2357.5472.9167.1758.77
2025.10
70.8821.9550.3871.6565.8256.14
2025.10
70.7520.853.0271.2764.5656.08
2025.10
69.8722.0350.3271.665.1955.8
2025.10
68.7320.847.1870.5764.854.42
2025.10
68.5619.8246.6170.3564.3353.93
2025.10
66.1220.3145.9769.2663.4653.02
2025.10
59.3419.9831.7164.1562.6747.57
2025.10
57.8722.1126.8663.662.1246.51
2025.10
56.6119.4926.4763.2260.8545.33
2025.10
54.7619.2521.6662.8959.7543.66
2025.10
50.8819.1616.5961.2158.8841.35
2025.10
50.2119.4916.1760.9458.6441.09
2025.10
49.2420.2316.4860.4558.4840.98
2025.10
47.1819.4916.1358.7657.6239.84
2025.10
46.319.4115.9359.2557.1439.61