Multiple Choice Question Answering on MMLU (Hendrycks Tasks) Part 2 benchmark leaderboard · SOTA2 Research

Multiple Choice Question Answering on MMLU (Hendrycks Tasks) Part 2

30.7High School Microeconomics AccuracyGPT-3 DaVinci

Updated 5mo ago

Evaluation Results

Method	Links
GPT-3 DaVinci2022.04		30.7	21.9	35.2	27.8	36.8	32.1	32.7	48.1	33.1	37	25.2	23.2	45.6	49.1	43	45	30.1	24.9	35.3	36.7	32.4	28.7	26.1	32.4	33.5	36.4	39.2	36.8	50	38.6	39.8
GPT-NeoX2022.04		29.4	29.8	28.3	31.9	30.9	29.5	22.4	40.5	29.8	25	22.7	26.8	28.2	32.1	34	29.9	28.9	23.2	37.9	29.3	27.2	23.4	26.7	28.7	27.5	34.5	37.6	28.4	36	36.1	38.6
GPT-J2022.04		27.7	27.2	27.3	29.2	28.9	28.3	26.5	39.7	26.4	27.8	29.4	22.3	23.3	30.3	31	27.5	28.3	23.7	34.6	26	24.4	26.2	24.1	27.6	28.4	28.2	36.3	27.9	34	35.5	33.3
GPT-3 Babbage2022.04		26.5	29.8	21.7	27.8	26	26.2	33.6	42.7	20.7	27.8	22.1	28.6	21.4	28.2	30	26.9	26.3	27.3	28.1	26.7	26.9	25.5	25.6	23.9	22.5	32.7	34.7	30.3	33	31.9	31
GPT-3 Curie2022.04		23.9	19.9	24.6	25.5	24	27	29.6	39.7	20.7	25.9	24.5	29.5	32	30.8	33	28.4	27.7	23.8	29.1	24.4	28.4	23.8	25.9	26.5	25.7	36.4	38.4	27.4	38	33.7	37.4
GPT-3 Ada2022.04		23.5	19.9	20.9	24.1	25.5	27.8	36.8	37.4	18.2	28.7	23.9	24.1	18.4	30.8	26	25.7	26.3	23.8	30.1	21.5	24.4	20.2	26.1	22.1	24.5	25.5	36.7	32.8	33	30.7	31.6