BIG-bench Hard (BBH) Comprehensive Evaluation benchmark leaderboard · SOTA2 Research

BIG-bench Hard (BBH) Comprehensive Evaluation

83.2Boolean ExpressionsPaLM540B

Evaluation Results

Method	Links
PaLM540B2023.03		83.2	61	53.6	60.8	28.4	53.6	37.6	70.8	60.4	87.2	1.6	62.4	51.2	44.5	38	76	48.8	78.1	80.4	39.6	19.6	51.2	32	62.7	40.9	52.3	—
GPT-NeoX2023.03		71.2	52.41	45.6	40.8	26	52.8	8	92	30.93	86.4	0.4	45.2	21.2	33.56	26	54	20.4	62.36	53.2	21.2	24.53	52.4	5.2	51.63	27.84	40.25	45
BLOOM176B2023.03		69.2	51.87	50	40.4	42	52.8	22.4	92	34	91.2	0	50	36.8	40.41	36.8	54.8	23.6	72.47	53.2	36.8	23.47	51.2	7.6	54.96	33.95	44.91	75
BLOOMBERGGPT2023.03		62.4	49.73	54.8	34	15.6	50.8	15.2	92	34.53	90.4	1.2	42	33.2	37.67	34.8	56	20	69.66	62.8	29.2	25.33	49.2	4.8	54.39	28.42	41.97	57
OPT66B2023.03		48.4	51.87	49.6	40.4	14.8	54	11.6	91.6	31.87	91.2	0.4	42	26	28.08	31.2	52.8	16.4	69.66	54.4	23.6	24	54	2.4	52.6	25.37	39.58	39