Our new X account is live! Follow @wizwand_team for updates

Commonsense knowledge probing on ViComTe (test)

49.6Color Spearman Correlation

GPT-J-6B

Updated 4d ago

Evaluation Results

Method	Links
GPT-J-6B 2022.10		49.6	50.9	47.1	70	37.9	46.5
OPT-30B 2022.10		48.1	52.1	48.4	72.1	38.1	47.5
GPT-Neo-1.3B 2022.10		47.4	48.4	46.1	64.3	36.4	41.2
GPT-Neo-2.7B 2022.10		47	43.6	45.2	62.1	36.1	37.7
GPT-J-6B 2022.10		44.3	38	46.9	65.7	38.5	42.3
OPT-30B 2022.10		41.9	41.3	46.4	60	38.3	44.7
RoBERTa-L-mnli 2022.10		41.4	38.2	43.5	67.9	31.6	37.7
GPT-Neo-2.7B 2022.10		41.3	29.3	43.5	50	35.5	30.6
GPT-Neo-1.3B 2022.10		40.1	31.7	44.2	52.9	35.1	34.5
BART-L-mnli 2022.10		38.2	32.6	42	69.3	32.6	40.5
SBERT 2022.10		38.2	40.6	33.6	35	30.4	34.9
BERT-L 2022.10		37.6	30.3	42.7	28.4	36.6	35.7
Z-LaVI w/o LM 2022.10		37.2	39.4	33.8	32.1	24.9	32.7
SimCSE 2022.10		36.3	40.4	34.7	40.7	29.2	33.5
RoBERTa-L-mnli 2022.10		35.3	25.1	41.7	66.4	28.3	34.9
BART-L-mnli 2022.10		34.6	27.5	41.5	68.6	30.5	37
Oscar-L 2022.10		31.8	17.1	40	38.1	39.2	40.5
SimCSE 2022.10		30.7	34.8	30.1	28.6	24.6	26.4
SBERT 2022.10		22.6	26.5	20.3	13.6	24.1	22.5
Random 2022.10		-0.1	6.6	1.3	7.1	-1.5	6