Large Language Model Reasoning and Understanding on Llama-3.2-3B-Instruct (test val) benchmark leaderboard · SOTA2 Research

Large Language Model Reasoning and Understanding on Llama-3.2-3B-Instruct (test val)

37.3MMLU ProBF16

Evaluation Results

Method	Links
BF162026.05		37.3	78.54	70.24	71.75	68.51	65.27	100
ScaleSweepMSE2026.05		35.09	74.68	65.62	70.26	68.75	62.88	96.34
ScaleSweep2026.05		34.97	73.69	70.61	70.37	66.46	63.22	96.86
4/62026.05		34.23	72.78	69.69	70.53	66.06	62.66	96
ScaleSweep2026.05		34.12	73.16	68.76	70.18	68.98	63.04	96.58
4/62026.05		34.1	71.95	67.28	70.11	65.51	61.79	94.67
ScaleSweepMSE2026.05		33.96	73.46	68.02	70.39	67.09	62.59	95.89
ScaleSweep2026.05		33.75	—	67.65	70.07	67.25	62.1	95.15
ScaleSweepMSE2026.05		33.44	72.33	65.62	70.22	66.93	61.71	94.54
AbsMax2026.05		33.43	71.49	63.96	70.3	68.27	61.49	94.21
AbsMax2026.05		33.29	71.57	65.43	69.85	68.75	61.78	94.65
ScaleSweep2026.05		33.1	72.33	67.84	70.08	67.8	62.23	95.34
ScaleSweepMSE2026.05		33.08	72.33	67.47	69.79	66.3	61.79	94.67
ScaleSweep2026.05		33.02	71.11	65.99	69.43	66.3	61.17	93.72
4/62026.05		32.41	71.11	66.17	69.5	67.4	61.32	93.95
ScaleSweepMSE2026.05		32.21	68.76	62.66	69.25	65.82	59.74	91.53
4/62026.05		31.99	69.83	67.84	69.63	66.69	61.19	93.76
ScaleSweep2026.05		31.93	69.9	66.91	69.73	66.38	60.97	93.41
ScaleSweepMSE2026.05		31.33	68.31	66.17	69.13	66.69	60.33	92.43
AbsMax2026.05		31.26	67.63	66.73	69.81	66.85	60.45	92.62
AbsMax2026.05		30.93	67.17	66.36	69.74	68.03	60.45	92.61
4/62026.05		30.73	66.11	65.8	69.01	64.8	59.29	90.84
4/62026.05		30.44	67.32	60.44	69.46	67.32	59	90.39
AbsMax2026.05		29.81	62.62	61.74	69.05	65.35	57.71	88.43
AbsMax2026.05		28	61.87	63.96	68.71	65.43	57.59	88.24