| Task , Accuracy , Centered | |
| hellaswag_zeroshot , 0.445031 , 0.260041 | |
| jeopardy , 0.116202 , 0.116202 | |
| bigbench_qa_wikidata , 0.524039 , 0.524039 | |
| arc_easy , 0.651936 , 0.535915 | |
| arc_challenge , 0.353242 , 0.137656 | |
| copa , 0.640000 , 0.280000 | |
| commonsense_qa , 0.325962 , 0.157453 | |
| piqa , 0.686072 , 0.372144 | |
| openbook_qa , 0.332000 , 0.109333 | |
| lambada_openai , 0.368523 , 0.368523 | |
| hellaswag , 0.445927 , 0.261236 | |
| winograd , 0.663004 , 0.326007 | |
| winogrande , 0.546961 , 0.093923 | |
| bigbench_dyck_languages , 0.126000 , 0.126000 | |
| agi_eval_lsat_ar , 0.269565 , 0.086956 | |
| bigbench_cs_algorithms , 0.342424 , 0.342424 | |
| bigbench_operators , 0.161905 , 0.161905 | |
| bigbench_repeat_copy_logic , 0.000000 , 0.000000 | |
| squad , 0.237938 , 0.237938 | |
| coqa , 0.200676 , 0.200676 | |
| boolq , 0.490214 , -0.341542 | |
| bigbench_language_identification , 0.257400 , 0.183058 | |
| CORE , , 0.206359 | |