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

Zero-shot Accuracy on Downstream Suite (BoolQ, HellaSwag, PIQA, RACE, WinoGrande)

82.4BoolQ Accuracy

Baseline

61.49666.92372.3577.777Oct 7, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.10
82.460.279.546.871.968.2
2024.10
79.257.178.14469.565.6
2024.10
78.959.98046.472.867.6
2024.10
78.360.579.54372.866.8
2024.10
78.362.181.14775.268.7
2024.10
78.257.97843.770.665.7
2024.10
77.95777.842.469.965
2024.10
77.157.178.742.769.265
2024.10
76.657.379.540.270.264.8
2024.10
76.457.279.340.770.564.8
2024.10
75.756.27841.569.164.1
2024.10
72.255.879.340.169.763.4
2024.10
72.25276.53863.560.4
2024.10
70.355.378.940.769.362.9
2024.10
6954.278.141.467.862.1
2024.10
66.645.17336.463.955.6
2024.10
66.552.377.638.966.960.4
2024.10
64.343.574.237.658.155.5
2024.10
62.339.971.634.558.753.4