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

Multiple-choice QA on PIQA, OpenBookQA, HellaSwag, and ARC

68.39PIQA Accuracy

NexusFormer

50.117254.861159.60564.3489Apr 21, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
68.3933.64559.6730.7247.48
2026.04
66.533439.9351.5127.943.97
2026.04
66.273339.8950.828.6743.73
2026.04
65.9932.639.9452.3528.6743.91
2026.04
65.232.638.446.827.242.04
2026.04
64.4730.641.4150.9727.7343.04
2026.04
64.4232.435.5349.3326.8841.71
2026.04
60.342929.7244.7424.8337.73
2026.04
59.7430.432.1842.9725.3438.13
2026.04
56.9128.831.5340.3625.9336.71
2026.04
53.322526.4234.8922.4432.41
2026.04
51.92826.2727.5726.1931.99
2026.04
51.412526.2632.8323.8131.86
2026.04
51.1425.425.928.224.230.97
2026.04
51.0327.427.2829.2924.1431.83
2026.04
50.822625.527.5726.8831.35