List your company

Language Modeling on SNLG and SNLU evaluation suites (test)

78.71SNLG ScoreTeacher baseline (OPT-6.7B)
48.934856.664964.39572.1251Feb 19, 2024

Evaluation results

MethodLinks
Teacher baseline (OPT-6.7B)Model Role=Teacher, Backbone=OPT-6.7B2024.0278.7123.4351.07
Teacher baseline (OPT-2.7B)Model Role=Teacher, Backbone=OPT-2.7B2024.0274.2122.2848.25
Teacher baseline (OPT-1.3B)Model Role=Teacher, Backbone=OPT-1.3B2024.0268.922.645.75
ImitKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0262.6619.5641.11
GKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0262.6219.2640.94
GKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0262.4619.2140.84
GKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0261.3619.0740.21
f-distill + ATKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0261.319.5440.42
f-distill + ATKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0261.2519.1840.22
GKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0261.2417.4839.36
GKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0261.2318.7740
ImitKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0260.7719.1839.97
ImitKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0260.7619.2940.02
f-distill + ATKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0260.719.0239.86
GKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0260.5916.8738.73
f-distillStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0260.0117.0838.55
ImitKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0259.8817.4638.67
ImitKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0259.8718.4139.14
f-distillStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0259.7419.4639.6
f-distillStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0259.0217.838.41
ImitKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0258.8617.2838.07
Teacher baseline (OPT-350M)Model Role=Teacher, Backbone=OPT-350M2024.0258.3320.3639.35
Supervised KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0257.5619.3138.43
Supervised KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0257.2619.4838.37
Supervised KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0256.7619.7338.25
Supervised KDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0255.5717.9936.78
Supervised KDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0255.4518.3336.89
Supervised KDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0255.318.6937
Reverse KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0254.419.436.9
Reverse KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0254.3719.1636.76
Reverse KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0254.3419.2736.8
f-distill + ATKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0252.6918.835.75
ImitKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0252.3618.6635.51
ImitKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0252.2718.3535.31
f-distillStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0252.1818.5735.37
Supervised KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0252.1619.5835.87
GKD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0251.918.5235.21
GKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0251.8717.3234.59
Reverse KDStudent Model=OPT-125M, Teacher Model=OPT-1.3B2024.0251.618.1534.87
Reverse KDStudent Model=OPT-125M, Teacher Model=OPT-2.7B2024.0251.2618.5634.91
Reverse KD + ATKDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0250.8619.1334.99
Supervised KDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0250.6218.8834.75
Reverse KDStudent Model=OPT-125M, Teacher Model=OPT-350M2024.0250.5418.0534.3
Reverse KDStudent Model=OPT-125M, Teacher Model=OPT-6.7B2024.0250.0818.3334.2