Zero-shot Reasoning on (ARC-C, BoolQ, HellaSwag, MathQA, MMLU, Race) benchmark leaderboard · SOTA2 Research

Zero-shot Reasoning on (ARC-C, BoolQ, HellaSwag, MathQA, MMLU, Race)

50.43ARC-C AccuracyBase

Evaluation Results

Method	Links
Base2025.10		50.43	83.61	81.08	35.71	59.64	40.86	58.56
Base2025.10		43.34	77.71	76	28.07	41.82	39.52	51.08
PGSVD2025.10		37.71	66.21	62.26	29.25	28.5	36.94	43.48
PGSVD2025.10		36.52	67.43	60.96	25.76	27.48	35.02	42.2
SVD-ALS2025.10		36.35	65.81	59.74	28.38	27.99	35.5	42.3
SVD-LLM2025.10		35.15	67.74	59.6	28.54	28.53	35.5	42.51
SVD-LLM2025.10		34.39	63.09	59.19	25.33	26.2	34.55	40.46
SVD-ALS2025.10		33.7	68.01	59.55	24.96	26.64	33.97	41.14
PGSVD2025.10		23.63	56.12	43.42	22.65	23.29	28.61	32.62
SVD-LLM2025.10		22.95	43.21	41.73	22.24	23.34	28.8	30.83
SVD-ALS2025.10		22.87	42.57	41.74	22.38	23.54	27.85	30.16
SVD-LLM2025.10		21.67	38.62	37.87	22.18	23.38	27.08	28.47
PGSVD2025.10		21.42	39.02	39.12	22.35	23.44	27.46	28.8
SVD-ALS2025.10		20.82	38.41	37.87	22.75	23.28	27.18	28.39