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

Zero-shot Commonsense Reasoning on Reasoning Suite (ARC-E, BoolQ, HSwag, LAMBADA, OBQA, PIQA, SocIQA, WinoGr.)

45.5ARC-E Accuracy

PathMoE

40.289641.642342.99544.3477Mar 18, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.03
45.555.4447.8247.8932.466.2138.4953.4348.4
2026.03
44.760.447.9549.4531.666.3240.9455.6449.62
2026.03
44.5756.4545.9946.1929.866.8738.8451.5447.53
2026.03
44.3659.9746.3346.1131.866.6539.252.9648.42
2026.03
44.0754.546.0547.1230.867.7438.7953.6747.84
2026.03
43.7359.9746.6546.9231.466.8140.6954.9348.89
2026.03
43.2760.5546.2645.2232.267.6339.6152.848.44
2026.03
42.5557.846.3646.7331.265.6737.5654.0647.74
2026.03
40.4961.137.9538.8328.264.4738.8950.9945.12