417 lines
14 KiB
JSON
417 lines
14 KiB
JSON
{
|
|
"results": {
|
|
"hendrycksTest-abstract_algebra": {
|
|
"acc": 0.27,
|
|
"acc_stderr": 0.04461960433384741,
|
|
"acc_norm": 0.27,
|
|
"acc_norm_stderr": 0.04461960433384741
|
|
},
|
|
"hendrycksTest-anatomy": {
|
|
"acc": 0.3925925925925926,
|
|
"acc_stderr": 0.04218506215368879,
|
|
"acc_norm": 0.3925925925925926,
|
|
"acc_norm_stderr": 0.04218506215368879
|
|
},
|
|
"hendrycksTest-astronomy": {
|
|
"acc": 0.3815789473684211,
|
|
"acc_stderr": 0.03953173377749194,
|
|
"acc_norm": 0.3815789473684211,
|
|
"acc_norm_stderr": 0.03953173377749194
|
|
},
|
|
"hendrycksTest-business_ethics": {
|
|
"acc": 0.4,
|
|
"acc_stderr": 0.049236596391733084,
|
|
"acc_norm": 0.4,
|
|
"acc_norm_stderr": 0.049236596391733084
|
|
},
|
|
"hendrycksTest-clinical_knowledge": {
|
|
"acc": 0.37735849056603776,
|
|
"acc_stderr": 0.029832808114796005,
|
|
"acc_norm": 0.37735849056603776,
|
|
"acc_norm_stderr": 0.029832808114796005
|
|
},
|
|
"hendrycksTest-college_biology": {
|
|
"acc": 0.3680555555555556,
|
|
"acc_stderr": 0.040329990539607195,
|
|
"acc_norm": 0.3680555555555556,
|
|
"acc_norm_stderr": 0.040329990539607195
|
|
},
|
|
"hendrycksTest-college_chemistry": {
|
|
"acc": 0.25,
|
|
"acc_stderr": 0.04351941398892446,
|
|
"acc_norm": 0.25,
|
|
"acc_norm_stderr": 0.04351941398892446
|
|
},
|
|
"hendrycksTest-college_computer_science": {
|
|
"acc": 0.37,
|
|
"acc_stderr": 0.04852365870939099,
|
|
"acc_norm": 0.37,
|
|
"acc_norm_stderr": 0.04852365870939099
|
|
},
|
|
"hendrycksTest-college_mathematics": {
|
|
"acc": 0.3,
|
|
"acc_stderr": 0.046056618647183814,
|
|
"acc_norm": 0.3,
|
|
"acc_norm_stderr": 0.046056618647183814
|
|
},
|
|
"hendrycksTest-college_medicine": {
|
|
"acc": 0.3236994219653179,
|
|
"acc_stderr": 0.0356760379963917,
|
|
"acc_norm": 0.3236994219653179,
|
|
"acc_norm_stderr": 0.0356760379963917
|
|
},
|
|
"hendrycksTest-college_physics": {
|
|
"acc": 0.22549019607843138,
|
|
"acc_stderr": 0.04158307533083286,
|
|
"acc_norm": 0.22549019607843138,
|
|
"acc_norm_stderr": 0.04158307533083286
|
|
},
|
|
"hendrycksTest-computer_security": {
|
|
"acc": 0.46,
|
|
"acc_stderr": 0.05009082659620332,
|
|
"acc_norm": 0.46,
|
|
"acc_norm_stderr": 0.05009082659620332
|
|
},
|
|
"hendrycksTest-conceptual_physics": {
|
|
"acc": 0.33617021276595743,
|
|
"acc_stderr": 0.03088161852067694,
|
|
"acc_norm": 0.33617021276595743,
|
|
"acc_norm_stderr": 0.03088161852067694
|
|
},
|
|
"hendrycksTest-econometrics": {
|
|
"acc": 0.3157894736842105,
|
|
"acc_stderr": 0.04372748290278007,
|
|
"acc_norm": 0.3157894736842105,
|
|
"acc_norm_stderr": 0.04372748290278007
|
|
},
|
|
"hendrycksTest-electrical_engineering": {
|
|
"acc": 0.3793103448275862,
|
|
"acc_stderr": 0.04043461861916747,
|
|
"acc_norm": 0.3793103448275862,
|
|
"acc_norm_stderr": 0.04043461861916747
|
|
},
|
|
"hendrycksTest-elementary_mathematics": {
|
|
"acc": 0.25925925925925924,
|
|
"acc_stderr": 0.02256989707491841,
|
|
"acc_norm": 0.25925925925925924,
|
|
"acc_norm_stderr": 0.02256989707491841
|
|
},
|
|
"hendrycksTest-formal_logic": {
|
|
"acc": 0.20634920634920634,
|
|
"acc_stderr": 0.03619604524124251,
|
|
"acc_norm": 0.20634920634920634,
|
|
"acc_norm_stderr": 0.03619604524124251
|
|
},
|
|
"hendrycksTest-global_facts": {
|
|
"acc": 0.3,
|
|
"acc_stderr": 0.046056618647183814,
|
|
"acc_norm": 0.3,
|
|
"acc_norm_stderr": 0.046056618647183814
|
|
},
|
|
"hendrycksTest-high_school_biology": {
|
|
"acc": 0.34838709677419355,
|
|
"acc_stderr": 0.02710482632810094,
|
|
"acc_norm": 0.34838709677419355,
|
|
"acc_norm_stderr": 0.02710482632810094
|
|
},
|
|
"hendrycksTest-high_school_chemistry": {
|
|
"acc": 0.3054187192118227,
|
|
"acc_stderr": 0.03240661565868408,
|
|
"acc_norm": 0.3054187192118227,
|
|
"acc_norm_stderr": 0.03240661565868408
|
|
},
|
|
"hendrycksTest-high_school_computer_science": {
|
|
"acc": 0.35,
|
|
"acc_stderr": 0.047937248544110196,
|
|
"acc_norm": 0.35,
|
|
"acc_norm_stderr": 0.047937248544110196
|
|
},
|
|
"hendrycksTest-high_school_european_history": {
|
|
"acc": 0.36363636363636365,
|
|
"acc_stderr": 0.03756335775187896,
|
|
"acc_norm": 0.36363636363636365,
|
|
"acc_norm_stderr": 0.03756335775187896
|
|
},
|
|
"hendrycksTest-high_school_geography": {
|
|
"acc": 0.3686868686868687,
|
|
"acc_stderr": 0.03437305501980619,
|
|
"acc_norm": 0.3686868686868687,
|
|
"acc_norm_stderr": 0.03437305501980619
|
|
},
|
|
"hendrycksTest-high_school_government_and_politics": {
|
|
"acc": 0.5181347150259067,
|
|
"acc_stderr": 0.036060650018329185,
|
|
"acc_norm": 0.5181347150259067,
|
|
"acc_norm_stderr": 0.036060650018329185
|
|
},
|
|
"hendrycksTest-high_school_macroeconomics": {
|
|
"acc": 0.34102564102564104,
|
|
"acc_stderr": 0.024035489676335068,
|
|
"acc_norm": 0.34102564102564104,
|
|
"acc_norm_stderr": 0.024035489676335068
|
|
},
|
|
"hendrycksTest-high_school_mathematics": {
|
|
"acc": 0.2851851851851852,
|
|
"acc_stderr": 0.027528599210340496,
|
|
"acc_norm": 0.2851851851851852,
|
|
"acc_norm_stderr": 0.027528599210340496
|
|
},
|
|
"hendrycksTest-high_school_microeconomics": {
|
|
"acc": 0.35294117647058826,
|
|
"acc_stderr": 0.03104194130405927,
|
|
"acc_norm": 0.35294117647058826,
|
|
"acc_norm_stderr": 0.03104194130405927
|
|
},
|
|
"hendrycksTest-high_school_physics": {
|
|
"acc": 0.31125827814569534,
|
|
"acc_stderr": 0.03780445850526732,
|
|
"acc_norm": 0.31125827814569534,
|
|
"acc_norm_stderr": 0.03780445850526732
|
|
},
|
|
"hendrycksTest-high_school_psychology": {
|
|
"acc": 0.4018348623853211,
|
|
"acc_stderr": 0.02102010617299701,
|
|
"acc_norm": 0.4018348623853211,
|
|
"acc_norm_stderr": 0.02102010617299701
|
|
},
|
|
"hendrycksTest-high_school_statistics": {
|
|
"acc": 0.3148148148148148,
|
|
"acc_stderr": 0.03167468706828979,
|
|
"acc_norm": 0.3148148148148148,
|
|
"acc_norm_stderr": 0.03167468706828979
|
|
},
|
|
"hendrycksTest-high_school_us_history": {
|
|
"acc": 0.37254901960784315,
|
|
"acc_stderr": 0.033933885849584025,
|
|
"acc_norm": 0.37254901960784315,
|
|
"acc_norm_stderr": 0.033933885849584025
|
|
},
|
|
"hendrycksTest-high_school_world_history": {
|
|
"acc": 0.4345991561181435,
|
|
"acc_stderr": 0.03226759995510144,
|
|
"acc_norm": 0.4345991561181435,
|
|
"acc_norm_stderr": 0.03226759995510144
|
|
},
|
|
"hendrycksTest-human_aging": {
|
|
"acc": 0.42152466367713004,
|
|
"acc_stderr": 0.03314190222110658,
|
|
"acc_norm": 0.42152466367713004,
|
|
"acc_norm_stderr": 0.03314190222110658
|
|
},
|
|
"hendrycksTest-human_sexuality": {
|
|
"acc": 0.3969465648854962,
|
|
"acc_stderr": 0.04291135671009224,
|
|
"acc_norm": 0.3969465648854962,
|
|
"acc_norm_stderr": 0.04291135671009224
|
|
},
|
|
"hendrycksTest-international_law": {
|
|
"acc": 0.5371900826446281,
|
|
"acc_stderr": 0.04551711196104218,
|
|
"acc_norm": 0.5371900826446281,
|
|
"acc_norm_stderr": 0.04551711196104218
|
|
},
|
|
"hendrycksTest-jurisprudence": {
|
|
"acc": 0.35185185185185186,
|
|
"acc_stderr": 0.046166311118017146,
|
|
"acc_norm": 0.35185185185185186,
|
|
"acc_norm_stderr": 0.046166311118017146
|
|
},
|
|
"hendrycksTest-logical_fallacies": {
|
|
"acc": 0.3374233128834356,
|
|
"acc_stderr": 0.037149084099355745,
|
|
"acc_norm": 0.3374233128834356,
|
|
"acc_norm_stderr": 0.037149084099355745
|
|
},
|
|
"hendrycksTest-machine_learning": {
|
|
"acc": 0.2857142857142857,
|
|
"acc_stderr": 0.042878587513404544,
|
|
"acc_norm": 0.2857142857142857,
|
|
"acc_norm_stderr": 0.042878587513404544
|
|
},
|
|
"hendrycksTest-management": {
|
|
"acc": 0.36893203883495146,
|
|
"acc_stderr": 0.04777615181156739,
|
|
"acc_norm": 0.36893203883495146,
|
|
"acc_norm_stderr": 0.04777615181156739
|
|
},
|
|
"hendrycksTest-marketing": {
|
|
"acc": 0.45726495726495725,
|
|
"acc_stderr": 0.03263622596380688,
|
|
"acc_norm": 0.45726495726495725,
|
|
"acc_norm_stderr": 0.03263622596380688
|
|
},
|
|
"hendrycksTest-medical_genetics": {
|
|
"acc": 0.44,
|
|
"acc_stderr": 0.04988876515698589,
|
|
"acc_norm": 0.44,
|
|
"acc_norm_stderr": 0.04988876515698589
|
|
},
|
|
"hendrycksTest-miscellaneous": {
|
|
"acc": 0.4763729246487867,
|
|
"acc_stderr": 0.01785998976517645,
|
|
"acc_norm": 0.4763729246487867,
|
|
"acc_norm_stderr": 0.01785998976517645
|
|
},
|
|
"hendrycksTest-moral_disputes": {
|
|
"acc": 0.3699421965317919,
|
|
"acc_stderr": 0.025992472029306386,
|
|
"acc_norm": 0.3699421965317919,
|
|
"acc_norm_stderr": 0.025992472029306386
|
|
},
|
|
"hendrycksTest-moral_scenarios": {
|
|
"acc": 0.23798882681564246,
|
|
"acc_stderr": 0.01424263007057488,
|
|
"acc_norm": 0.23798882681564246,
|
|
"acc_norm_stderr": 0.01424263007057488
|
|
},
|
|
"hendrycksTest-nutrition": {
|
|
"acc": 0.4150326797385621,
|
|
"acc_stderr": 0.028213504177824103,
|
|
"acc_norm": 0.4150326797385621,
|
|
"acc_norm_stderr": 0.028213504177824103
|
|
},
|
|
"hendrycksTest-philosophy": {
|
|
"acc": 0.4405144694533762,
|
|
"acc_stderr": 0.028196400574197426,
|
|
"acc_norm": 0.4405144694533762,
|
|
"acc_norm_stderr": 0.028196400574197426
|
|
},
|
|
"hendrycksTest-prehistory": {
|
|
"acc": 0.37962962962962965,
|
|
"acc_stderr": 0.02700252103451648,
|
|
"acc_norm": 0.37962962962962965,
|
|
"acc_norm_stderr": 0.02700252103451648
|
|
},
|
|
"hendrycksTest-professional_accounting": {
|
|
"acc": 0.29432624113475175,
|
|
"acc_stderr": 0.02718712701150379,
|
|
"acc_norm": 0.29432624113475175,
|
|
"acc_norm_stderr": 0.02718712701150379
|
|
},
|
|
"hendrycksTest-professional_law": {
|
|
"acc": 0.30378096479791394,
|
|
"acc_stderr": 0.011745787720472472,
|
|
"acc_norm": 0.30378096479791394,
|
|
"acc_norm_stderr": 0.011745787720472472
|
|
},
|
|
"hendrycksTest-professional_medicine": {
|
|
"acc": 0.40808823529411764,
|
|
"acc_stderr": 0.029855261393483924,
|
|
"acc_norm": 0.40808823529411764,
|
|
"acc_norm_stderr": 0.029855261393483924
|
|
},
|
|
"hendrycksTest-professional_psychology": {
|
|
"acc": 0.35294117647058826,
|
|
"acc_stderr": 0.01933314202079706,
|
|
"acc_norm": 0.35294117647058826,
|
|
"acc_norm_stderr": 0.01933314202079706
|
|
},
|
|
"hendrycksTest-public_relations": {
|
|
"acc": 0.4090909090909091,
|
|
"acc_stderr": 0.04709306978661896,
|
|
"acc_norm": 0.4090909090909091,
|
|
"acc_norm_stderr": 0.04709306978661896
|
|
},
|
|
"hendrycksTest-security_studies": {
|
|
"acc": 0.3224489795918367,
|
|
"acc_stderr": 0.029923100563683906,
|
|
"acc_norm": 0.3224489795918367,
|
|
"acc_norm_stderr": 0.029923100563683906
|
|
},
|
|
"hendrycksTest-sociology": {
|
|
"acc": 0.42786069651741293,
|
|
"acc_stderr": 0.03498541988407795,
|
|
"acc_norm": 0.42786069651741293,
|
|
"acc_norm_stderr": 0.03498541988407795
|
|
},
|
|
"hendrycksTest-us_foreign_policy": {
|
|
"acc": 0.53,
|
|
"acc_stderr": 0.05016135580465919,
|
|
"acc_norm": 0.53,
|
|
"acc_norm_stderr": 0.05016135580465919
|
|
},
|
|
"hendrycksTest-virology": {
|
|
"acc": 0.3614457831325301,
|
|
"acc_stderr": 0.037400593820293204,
|
|
"acc_norm": 0.3614457831325301,
|
|
"acc_norm_stderr": 0.037400593820293204
|
|
},
|
|
"hendrycksTest-world_religions": {
|
|
"acc": 0.4678362573099415,
|
|
"acc_stderr": 0.038268824176603704,
|
|
"acc_norm": 0.4678362573099415,
|
|
"acc_norm_stderr": 0.038268824176603704
|
|
}
|
|
},
|
|
"versions": {
|
|
"hendrycksTest-abstract_algebra": 1,
|
|
"hendrycksTest-anatomy": 1,
|
|
"hendrycksTest-astronomy": 1,
|
|
"hendrycksTest-business_ethics": 1,
|
|
"hendrycksTest-clinical_knowledge": 1,
|
|
"hendrycksTest-college_biology": 1,
|
|
"hendrycksTest-college_chemistry": 1,
|
|
"hendrycksTest-college_computer_science": 1,
|
|
"hendrycksTest-college_mathematics": 1,
|
|
"hendrycksTest-college_medicine": 1,
|
|
"hendrycksTest-college_physics": 1,
|
|
"hendrycksTest-computer_security": 1,
|
|
"hendrycksTest-conceptual_physics": 1,
|
|
"hendrycksTest-econometrics": 1,
|
|
"hendrycksTest-electrical_engineering": 1,
|
|
"hendrycksTest-elementary_mathematics": 1,
|
|
"hendrycksTest-formal_logic": 1,
|
|
"hendrycksTest-global_facts": 1,
|
|
"hendrycksTest-high_school_biology": 1,
|
|
"hendrycksTest-high_school_chemistry": 1,
|
|
"hendrycksTest-high_school_computer_science": 1,
|
|
"hendrycksTest-high_school_european_history": 1,
|
|
"hendrycksTest-high_school_geography": 1,
|
|
"hendrycksTest-high_school_government_and_politics": 1,
|
|
"hendrycksTest-high_school_macroeconomics": 1,
|
|
"hendrycksTest-high_school_mathematics": 1,
|
|
"hendrycksTest-high_school_microeconomics": 1,
|
|
"hendrycksTest-high_school_physics": 1,
|
|
"hendrycksTest-high_school_psychology": 1,
|
|
"hendrycksTest-high_school_statistics": 1,
|
|
"hendrycksTest-high_school_us_history": 1,
|
|
"hendrycksTest-high_school_world_history": 1,
|
|
"hendrycksTest-human_aging": 1,
|
|
"hendrycksTest-human_sexuality": 1,
|
|
"hendrycksTest-international_law": 1,
|
|
"hendrycksTest-jurisprudence": 1,
|
|
"hendrycksTest-logical_fallacies": 1,
|
|
"hendrycksTest-machine_learning": 1,
|
|
"hendrycksTest-management": 1,
|
|
"hendrycksTest-marketing": 1,
|
|
"hendrycksTest-medical_genetics": 1,
|
|
"hendrycksTest-miscellaneous": 1,
|
|
"hendrycksTest-moral_disputes": 1,
|
|
"hendrycksTest-moral_scenarios": 1,
|
|
"hendrycksTest-nutrition": 1,
|
|
"hendrycksTest-philosophy": 1,
|
|
"hendrycksTest-prehistory": 1,
|
|
"hendrycksTest-professional_accounting": 1,
|
|
"hendrycksTest-professional_law": 1,
|
|
"hendrycksTest-professional_medicine": 1,
|
|
"hendrycksTest-professional_psychology": 1,
|
|
"hendrycksTest-public_relations": 1,
|
|
"hendrycksTest-security_studies": 1,
|
|
"hendrycksTest-sociology": 1,
|
|
"hendrycksTest-us_foreign_policy": 1,
|
|
"hendrycksTest-virology": 1,
|
|
"hendrycksTest-world_religions": 1
|
|
},
|
|
"config": {
|
|
"model": "sparseml",
|
|
"model_args": "pretrained=/network/alexandre/research/cerebras/llama2_7B_sparse70_retrained/checkpoint,dtype=bfloat16",
|
|
"num_fewshot": 5,
|
|
"batch_size": "4",
|
|
"batch_sizes": [],
|
|
"device": "cuda:6",
|
|
"no_cache": true,
|
|
"limit": null,
|
|
"bootstrap_iters": 100000,
|
|
"description_dict": {}
|
|
}
|
|
} |