417 lines
14 KiB
JSON
417 lines
14 KiB
JSON
{
|
|
"results": {
|
|
"hendrycksTest-abstract_algebra": {
|
|
"acc": 0.34,
|
|
"acc_stderr": 0.047609522856952365,
|
|
"acc_norm": 0.34,
|
|
"acc_norm_stderr": 0.047609522856952365
|
|
},
|
|
"hendrycksTest-anatomy": {
|
|
"acc": 0.5037037037037037,
|
|
"acc_stderr": 0.04319223625811331,
|
|
"acc_norm": 0.5037037037037037,
|
|
"acc_norm_stderr": 0.04319223625811331
|
|
},
|
|
"hendrycksTest-astronomy": {
|
|
"acc": 0.4276315789473684,
|
|
"acc_stderr": 0.04026097083296558,
|
|
"acc_norm": 0.4276315789473684,
|
|
"acc_norm_stderr": 0.04026097083296558
|
|
},
|
|
"hendrycksTest-business_ethics": {
|
|
"acc": 0.54,
|
|
"acc_stderr": 0.05009082659620333,
|
|
"acc_norm": 0.54,
|
|
"acc_norm_stderr": 0.05009082659620333
|
|
},
|
|
"hendrycksTest-clinical_knowledge": {
|
|
"acc": 0.4679245283018868,
|
|
"acc_stderr": 0.030709486992556545,
|
|
"acc_norm": 0.4679245283018868,
|
|
"acc_norm_stderr": 0.030709486992556545
|
|
},
|
|
"hendrycksTest-college_biology": {
|
|
"acc": 0.5,
|
|
"acc_stderr": 0.04181210050035455,
|
|
"acc_norm": 0.5,
|
|
"acc_norm_stderr": 0.04181210050035455
|
|
},
|
|
"hendrycksTest-college_chemistry": {
|
|
"acc": 0.29,
|
|
"acc_stderr": 0.045604802157206845,
|
|
"acc_norm": 0.29,
|
|
"acc_norm_stderr": 0.045604802157206845
|
|
},
|
|
"hendrycksTest-college_computer_science": {
|
|
"acc": 0.4,
|
|
"acc_stderr": 0.04923659639173309,
|
|
"acc_norm": 0.4,
|
|
"acc_norm_stderr": 0.04923659639173309
|
|
},
|
|
"hendrycksTest-college_mathematics": {
|
|
"acc": 0.31,
|
|
"acc_stderr": 0.04648231987117316,
|
|
"acc_norm": 0.31,
|
|
"acc_norm_stderr": 0.04648231987117316
|
|
},
|
|
"hendrycksTest-college_medicine": {
|
|
"acc": 0.4393063583815029,
|
|
"acc_stderr": 0.03784271932887467,
|
|
"acc_norm": 0.4393063583815029,
|
|
"acc_norm_stderr": 0.03784271932887467
|
|
},
|
|
"hendrycksTest-college_physics": {
|
|
"acc": 0.20588235294117646,
|
|
"acc_stderr": 0.04023382273617747,
|
|
"acc_norm": 0.20588235294117646,
|
|
"acc_norm_stderr": 0.04023382273617747
|
|
},
|
|
"hendrycksTest-computer_security": {
|
|
"acc": 0.61,
|
|
"acc_stderr": 0.04902071300001975,
|
|
"acc_norm": 0.61,
|
|
"acc_norm_stderr": 0.04902071300001975
|
|
},
|
|
"hendrycksTest-conceptual_physics": {
|
|
"acc": 0.43829787234042555,
|
|
"acc_stderr": 0.03243618636108101,
|
|
"acc_norm": 0.43829787234042555,
|
|
"acc_norm_stderr": 0.03243618636108101
|
|
},
|
|
"hendrycksTest-econometrics": {
|
|
"acc": 0.2894736842105263,
|
|
"acc_stderr": 0.04266339443159393,
|
|
"acc_norm": 0.2894736842105263,
|
|
"acc_norm_stderr": 0.04266339443159393
|
|
},
|
|
"hendrycksTest-electrical_engineering": {
|
|
"acc": 0.46206896551724136,
|
|
"acc_stderr": 0.041546596717075474,
|
|
"acc_norm": 0.46206896551724136,
|
|
"acc_norm_stderr": 0.041546596717075474
|
|
},
|
|
"hendrycksTest-elementary_mathematics": {
|
|
"acc": 0.2962962962962963,
|
|
"acc_stderr": 0.023517294335963276,
|
|
"acc_norm": 0.2962962962962963,
|
|
"acc_norm_stderr": 0.023517294335963276
|
|
},
|
|
"hendrycksTest-formal_logic": {
|
|
"acc": 0.30952380952380953,
|
|
"acc_stderr": 0.04134913018303316,
|
|
"acc_norm": 0.30952380952380953,
|
|
"acc_norm_stderr": 0.04134913018303316
|
|
},
|
|
"hendrycksTest-global_facts": {
|
|
"acc": 0.38,
|
|
"acc_stderr": 0.048783173121456316,
|
|
"acc_norm": 0.38,
|
|
"acc_norm_stderr": 0.048783173121456316
|
|
},
|
|
"hendrycksTest-high_school_biology": {
|
|
"acc": 0.5419354838709678,
|
|
"acc_stderr": 0.028343787250540615,
|
|
"acc_norm": 0.5419354838709678,
|
|
"acc_norm_stderr": 0.028343787250540615
|
|
},
|
|
"hendrycksTest-high_school_chemistry": {
|
|
"acc": 0.3448275862068966,
|
|
"acc_stderr": 0.03344283744280458,
|
|
"acc_norm": 0.3448275862068966,
|
|
"acc_norm_stderr": 0.03344283744280458
|
|
},
|
|
"hendrycksTest-high_school_computer_science": {
|
|
"acc": 0.4,
|
|
"acc_stderr": 0.049236596391733084,
|
|
"acc_norm": 0.4,
|
|
"acc_norm_stderr": 0.049236596391733084
|
|
},
|
|
"hendrycksTest-high_school_european_history": {
|
|
"acc": 0.6363636363636364,
|
|
"acc_stderr": 0.03756335775187897,
|
|
"acc_norm": 0.6363636363636364,
|
|
"acc_norm_stderr": 0.03756335775187897
|
|
},
|
|
"hendrycksTest-high_school_geography": {
|
|
"acc": 0.5353535353535354,
|
|
"acc_stderr": 0.03553436368828061,
|
|
"acc_norm": 0.5353535353535354,
|
|
"acc_norm_stderr": 0.03553436368828061
|
|
},
|
|
"hendrycksTest-high_school_government_and_politics": {
|
|
"acc": 0.7150259067357513,
|
|
"acc_stderr": 0.032577140777096614,
|
|
"acc_norm": 0.7150259067357513,
|
|
"acc_norm_stderr": 0.032577140777096614
|
|
},
|
|
"hendrycksTest-high_school_macroeconomics": {
|
|
"acc": 0.4564102564102564,
|
|
"acc_stderr": 0.025254485424799602,
|
|
"acc_norm": 0.4564102564102564,
|
|
"acc_norm_stderr": 0.025254485424799602
|
|
},
|
|
"hendrycksTest-high_school_mathematics": {
|
|
"acc": 0.2740740740740741,
|
|
"acc_stderr": 0.027195934804085622,
|
|
"acc_norm": 0.2740740740740741,
|
|
"acc_norm_stderr": 0.027195934804085622
|
|
},
|
|
"hendrycksTest-high_school_microeconomics": {
|
|
"acc": 0.4495798319327731,
|
|
"acc_stderr": 0.03231293497137707,
|
|
"acc_norm": 0.4495798319327731,
|
|
"acc_norm_stderr": 0.03231293497137707
|
|
},
|
|
"hendrycksTest-high_school_physics": {
|
|
"acc": 0.2847682119205298,
|
|
"acc_stderr": 0.03684881521389023,
|
|
"acc_norm": 0.2847682119205298,
|
|
"acc_norm_stderr": 0.03684881521389023
|
|
},
|
|
"hendrycksTest-high_school_psychology": {
|
|
"acc": 0.655045871559633,
|
|
"acc_stderr": 0.020380605405066955,
|
|
"acc_norm": 0.655045871559633,
|
|
"acc_norm_stderr": 0.020380605405066955
|
|
},
|
|
"hendrycksTest-high_school_statistics": {
|
|
"acc": 0.32407407407407407,
|
|
"acc_stderr": 0.03191923445686185,
|
|
"acc_norm": 0.32407407407407407,
|
|
"acc_norm_stderr": 0.03191923445686185
|
|
},
|
|
"hendrycksTest-high_school_us_history": {
|
|
"acc": 0.6274509803921569,
|
|
"acc_stderr": 0.03393388584958406,
|
|
"acc_norm": 0.6274509803921569,
|
|
"acc_norm_stderr": 0.03393388584958406
|
|
},
|
|
"hendrycksTest-high_school_world_history": {
|
|
"acc": 0.6497890295358649,
|
|
"acc_stderr": 0.031052391937584346,
|
|
"acc_norm": 0.6497890295358649,
|
|
"acc_norm_stderr": 0.031052391937584346
|
|
},
|
|
"hendrycksTest-human_aging": {
|
|
"acc": 0.547085201793722,
|
|
"acc_stderr": 0.033408675019233246,
|
|
"acc_norm": 0.547085201793722,
|
|
"acc_norm_stderr": 0.033408675019233246
|
|
},
|
|
"hendrycksTest-human_sexuality": {
|
|
"acc": 0.5648854961832062,
|
|
"acc_stderr": 0.04348208051644858,
|
|
"acc_norm": 0.5648854961832062,
|
|
"acc_norm_stderr": 0.04348208051644858
|
|
},
|
|
"hendrycksTest-international_law": {
|
|
"acc": 0.6363636363636364,
|
|
"acc_stderr": 0.043913262867240704,
|
|
"acc_norm": 0.6363636363636364,
|
|
"acc_norm_stderr": 0.043913262867240704
|
|
},
|
|
"hendrycksTest-jurisprudence": {
|
|
"acc": 0.5555555555555556,
|
|
"acc_stderr": 0.04803752235190193,
|
|
"acc_norm": 0.5555555555555556,
|
|
"acc_norm_stderr": 0.04803752235190193
|
|
},
|
|
"hendrycksTest-logical_fallacies": {
|
|
"acc": 0.49693251533742333,
|
|
"acc_stderr": 0.03928297078179663,
|
|
"acc_norm": 0.49693251533742333,
|
|
"acc_norm_stderr": 0.03928297078179663
|
|
},
|
|
"hendrycksTest-machine_learning": {
|
|
"acc": 0.39285714285714285,
|
|
"acc_stderr": 0.04635550135609976,
|
|
"acc_norm": 0.39285714285714285,
|
|
"acc_norm_stderr": 0.04635550135609976
|
|
},
|
|
"hendrycksTest-management": {
|
|
"acc": 0.5728155339805825,
|
|
"acc_stderr": 0.048979577377811674,
|
|
"acc_norm": 0.5728155339805825,
|
|
"acc_norm_stderr": 0.048979577377811674
|
|
},
|
|
"hendrycksTest-marketing": {
|
|
"acc": 0.6965811965811965,
|
|
"acc_stderr": 0.030118210106942638,
|
|
"acc_norm": 0.6965811965811965,
|
|
"acc_norm_stderr": 0.030118210106942638
|
|
},
|
|
"hendrycksTest-medical_genetics": {
|
|
"acc": 0.53,
|
|
"acc_stderr": 0.05016135580465919,
|
|
"acc_norm": 0.53,
|
|
"acc_norm_stderr": 0.05016135580465919
|
|
},
|
|
"hendrycksTest-miscellaneous": {
|
|
"acc": 0.6602809706257982,
|
|
"acc_stderr": 0.016936394114301635,
|
|
"acc_norm": 0.6602809706257982,
|
|
"acc_norm_stderr": 0.016936394114301635
|
|
},
|
|
"hendrycksTest-moral_disputes": {
|
|
"acc": 0.5317919075144508,
|
|
"acc_stderr": 0.026864624366756643,
|
|
"acc_norm": 0.5317919075144508,
|
|
"acc_norm_stderr": 0.026864624366756643
|
|
},
|
|
"hendrycksTest-moral_scenarios": {
|
|
"acc": 0.24581005586592178,
|
|
"acc_stderr": 0.014400296429225627,
|
|
"acc_norm": 0.24581005586592178,
|
|
"acc_norm_stderr": 0.014400296429225627
|
|
},
|
|
"hendrycksTest-nutrition": {
|
|
"acc": 0.5163398692810458,
|
|
"acc_stderr": 0.02861462475280544,
|
|
"acc_norm": 0.5163398692810458,
|
|
"acc_norm_stderr": 0.02861462475280544
|
|
},
|
|
"hendrycksTest-philosophy": {
|
|
"acc": 0.6012861736334405,
|
|
"acc_stderr": 0.027809322585774496,
|
|
"acc_norm": 0.6012861736334405,
|
|
"acc_norm_stderr": 0.027809322585774496
|
|
},
|
|
"hendrycksTest-prehistory": {
|
|
"acc": 0.5185185185185185,
|
|
"acc_stderr": 0.02780165621232366,
|
|
"acc_norm": 0.5185185185185185,
|
|
"acc_norm_stderr": 0.02780165621232366
|
|
},
|
|
"hendrycksTest-professional_accounting": {
|
|
"acc": 0.3723404255319149,
|
|
"acc_stderr": 0.028838921471251458,
|
|
"acc_norm": 0.3723404255319149,
|
|
"acc_norm_stderr": 0.028838921471251458
|
|
},
|
|
"hendrycksTest-professional_law": {
|
|
"acc": 0.3683181225554107,
|
|
"acc_stderr": 0.012319403369564637,
|
|
"acc_norm": 0.3683181225554107,
|
|
"acc_norm_stderr": 0.012319403369564637
|
|
},
|
|
"hendrycksTest-professional_medicine": {
|
|
"acc": 0.5330882352941176,
|
|
"acc_stderr": 0.030306257722468317,
|
|
"acc_norm": 0.5330882352941176,
|
|
"acc_norm_stderr": 0.030306257722468317
|
|
},
|
|
"hendrycksTest-professional_psychology": {
|
|
"acc": 0.46405228758169936,
|
|
"acc_stderr": 0.020175488765484036,
|
|
"acc_norm": 0.46405228758169936,
|
|
"acc_norm_stderr": 0.020175488765484036
|
|
},
|
|
"hendrycksTest-public_relations": {
|
|
"acc": 0.5545454545454546,
|
|
"acc_stderr": 0.047605488214603246,
|
|
"acc_norm": 0.5545454545454546,
|
|
"acc_norm_stderr": 0.047605488214603246
|
|
},
|
|
"hendrycksTest-security_studies": {
|
|
"acc": 0.4897959183673469,
|
|
"acc_stderr": 0.03200255347893782,
|
|
"acc_norm": 0.4897959183673469,
|
|
"acc_norm_stderr": 0.03200255347893782
|
|
},
|
|
"hendrycksTest-sociology": {
|
|
"acc": 0.6417910447761194,
|
|
"acc_stderr": 0.03390393042268814,
|
|
"acc_norm": 0.6417910447761194,
|
|
"acc_norm_stderr": 0.03390393042268814
|
|
},
|
|
"hendrycksTest-us_foreign_policy": {
|
|
"acc": 0.65,
|
|
"acc_stderr": 0.047937248544110196,
|
|
"acc_norm": 0.65,
|
|
"acc_norm_stderr": 0.047937248544110196
|
|
},
|
|
"hendrycksTest-virology": {
|
|
"acc": 0.39759036144578314,
|
|
"acc_stderr": 0.038099730845402184,
|
|
"acc_norm": 0.39759036144578314,
|
|
"acc_norm_stderr": 0.038099730845402184
|
|
},
|
|
"hendrycksTest-world_religions": {
|
|
"acc": 0.7134502923976608,
|
|
"acc_stderr": 0.03467826685703826,
|
|
"acc_norm": 0.7134502923976608,
|
|
"acc_norm_stderr": 0.03467826685703826
|
|
}
|
|
},
|
|
"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": "hf",
|
|
"model_args": "pretrained=/cache/shubhra/models/llama2_denseft_playtpus_dolphin_LR1e-5_E2_noGC_LRScosine/combined/,trust_remote_code=True,dtype=bfloat16",
|
|
"num_fewshot": 5,
|
|
"batch_size": "16",
|
|
"batch_sizes": [],
|
|
"device": "cuda:0",
|
|
"no_cache": true,
|
|
"limit": null,
|
|
"bootstrap_iters": 100000,
|
|
"description_dict": {}
|
|
}
|
|
} |