List your company

ResearchFormerly WizWand

Zero-shot Accuracy on Downstream Suite (BoolQ, HellaSwag, PIQA, RACE, WinoGrande)

82.4BoolQ AccuracyBaseline

Evaluation results

Method	Links
Baseline2024.10		82.4	60.2	79.5	46.8	71.9	68.2
Baseline2024.10		79.2	57.1	78.1	44	69.5	65.6
ESPACE2024.10		78.9	59.9	80	46.4	72.8	67.6
ESPACE2024.10		78.3	60.5	79.5	43	72.8	66.8
Baseline2024.10		78.3	62.1	81.1	47	75.2	68.7
Retrained2024.10		78.2	57.9	78	43.7	70.6	65.7
ESPACE2024.10		77.9	57	77.8	42.4	69.9	65
ESPACE2024.10		77.1	57.1	78.7	42.7	69.2	65
ESPACE2024.10		76.6	57.3	79.5	40.2	70.2	64.8
Baseline2024.10		76.4	57.2	79.3	40.7	70.5	64.8
ESPACE2024.10		75.7	56.2	78	41.5	69.1	64.1
ESPACE2024.10		72.2	55.8	79.3	40.1	69.7	63.4
ESPACE2024.10		72.2	52	76.5	38	63.5	60.4
ESPACE2024.10		70.3	55.3	78.9	40.7	69.3	62.9
Baseline2024.10		69	54.2	78.1	41.4	67.8	62.1
ESPACE2024.10		66.6	45.1	73	36.4	63.9	55.6
ESPACE2024.10		66.5	52.3	77.6	38.9	66.9	60.4
Baseline2024.10		64.3	43.5	74.2	37.6	58.1	55.5
ESPACE2024.10		62.3	39.9	71.6	34.5	58.7	53.4