mirror of
https://github.com/open-compass/opencompass.git
synced 2025-05-30 16:03:24 +08:00
89 lines
2.5 KiB
Python
89 lines
2.5 KiB
Python
supergpqa_summary_groups = []
|
|
|
|
# gaokao-bench
|
|
supergpqa_weights = {
|
|
'Electronic_Science_and_Technology': 246,
|
|
'Philosophy': 347,
|
|
'Traditional_Chinese_Medicine': 268,
|
|
'Applied_Economics': 723,
|
|
'Mathematics': 2622,
|
|
'Physics': 2845,
|
|
'Clinical_Medicine': 1218,
|
|
'Computer_Science_and_Technology': 763,
|
|
'Information_and_Communication_Engineering': 504,
|
|
'Control_Science_and_Engineering': 190,
|
|
'Theoretical_Economics': 150,
|
|
'Law': 591,
|
|
'History': 674,
|
|
'Basic_Medicine': 567,
|
|
'Education': 247,
|
|
'Materials_Science_and_Engineering': 289,
|
|
'Electrical_Engineering': 556,
|
|
'Systems_Science': 50,
|
|
'Power_Engineering_and_Engineering_Thermophysics': 684,
|
|
'Military_Science': 205,
|
|
'Biology': 1120,
|
|
'Business_Administration': 142,
|
|
'Language_and_Literature': 440,
|
|
'Public_Health_and_Preventive_Medicine': 292,
|
|
'Political_Science': 65,
|
|
'Chemistry': 1769,
|
|
'Hydraulic_Engineering': 218,
|
|
'Chemical_Engineering_and_Technology': 410,
|
|
'Pharmacy': 278,
|
|
'Geography': 133,
|
|
'Art_Studies': 603,
|
|
'Architecture': 162,
|
|
'Forestry_Engineering': 100,
|
|
'Public_Administration': 151,
|
|
'Oceanography': 200,
|
|
'Journalism_and_Communication': 207,
|
|
'Nuclear_Science_and_Technology': 107,
|
|
'Weapon_Science_and_Technology': 100,
|
|
'Naval_Architecture_and_Ocean_Engineering': 138,
|
|
'Environmental_Science_and_Engineering': 189,
|
|
'Transportation_Engineering': 251,
|
|
'Geology': 341,
|
|
'Physical_Oceanography': 50,
|
|
'Musicology': 426,
|
|
'Stomatology': 132,
|
|
'Aquaculture': 56,
|
|
'Mechanical_Engineering': 176,
|
|
'Aeronautical_and_Astronautical_Science_and_Technology': 119,
|
|
'Civil_Engineering': 358,
|
|
'Mechanics': 908,
|
|
'Petroleum_and_Natural_Gas_Engineering': 112,
|
|
'Sociology': 143,
|
|
'Food_Science_and_Engineering': 109,
|
|
'Agricultural_Engineering': 104,
|
|
'Surveying_and_Mapping_Science_and_Technology': 168,
|
|
'Metallurgical_Engineering': 255,
|
|
'Library,_Information_and_Archival_Management': 150,
|
|
'Mining_Engineering': 100,
|
|
'Astronomy': 405,
|
|
'Geological_Resources_and_Geological_Engineering': 50,
|
|
'Atmospheric_Science': 203,
|
|
'Optical_Engineering': 376,
|
|
'Animal_Husbandry': 103,
|
|
'Geophysics': 100,
|
|
'Crop_Science': 145,
|
|
'Management_Science_and_Engineering': 58,
|
|
'Psychology': 87,
|
|
'Forestry': 131,
|
|
'Textile_Science_and_Engineering': 100,
|
|
'Veterinary_Medicine': 50,
|
|
'Instrument_Science_and_Technology': 50,
|
|
'Physical_Education': 150,
|
|
}
|
|
supergpqa_weights = {
|
|
'supergpqa_' + k: v for k, v in supergpqa_weights.items()
|
|
}
|
|
supergpqa_summary_groups.append(
|
|
{
|
|
'name': 'SuperGPQA',
|
|
'subsets':[[k, 'accuracy'] for k in supergpqa_weights.keys()],
|
|
'weights': supergpqa_weights,
|
|
}
|
|
)
|
|
|
|
print(supergpqa_summary_groups) |