Mathematical Reasoning on AfriMGSM benchmark leaderboard · SOTA2 Research

Mathematical Reasoning on AfriMGSM

59.2Accuracy (Amharic)MERLIN

Evaluation Results

Method	Links
MERLIN2025.09		59.2	41.2	60.8	52	52	49.6	46.8	32	58.8	56	76.8	37.6	26	51.2	54	55.6	50.6
GPT-4o2025.09		57.6	8.8	64.8	57.6	60.4	51.2	51.6	61.2	58.4	60.8	78.8	31.2	28	52.4	62	57.2	52.6
SLAM2025.09		51.6	26	43.2	33.6	41.6	37.2	33.6	45.6	40	33.2	57.6	21.2	18.8	30.4	41.2	34.8	39
MERLIN2025.09		45.6	12.8	60.8	37.6	48.8	26.4	33.6	39.6	44	44.4	58.8	7.6	9.2	38	43.2	41.2	37
QAlign-LoRA2025.09		45.2	9.6	44.8	29.6	30.8	19.2	21.6	14.8	31.2	30.4	65.6	12.4	6.4	26.8	26	30.4	33.7
MindMerger2025.09		45.2	32.4	44.8	34.4	43.2	38.4	36.4	40	44.8	36.8	53.2	24.4	21.2	33.6	34.8	40.4	37.7
MindMerger2025.09		34.4	12	54.4	33.6	43.2	22	30	28.8	40.4	42.8	60	8.4	7.2	32	41.2	40.8	33.2
GPT-4o-mini2025.09		31.6	6	56	33.6	48	25.6	29.2	39.2	44.8	36.8	70.8	15.6	7.6	32.4	45.6	43.6	35.4
LangBridge2025.09		31.6	2	41.2	14	28.4	4	11.2	17.2	30	29.6	54.4	2.8	0.8	18.4	21.2	27.2	20.9
LayAlign2025.09		30	6.4	36.4	27.2	32	16	21.6	26.8	31.6	30	42.8	4	6	23.6	30.4	29.6	24.6
Gemma 2 9B2025.09		26.4	4.8	35.2	11.6	22.4	11.6	17.6	9.6	26	22	61.6	9.2	3.6	20.8	13.6	21.2	19.8
QAlign-LoRA2025.09		4.4	3.6	3.2	1.6	5.6	7.6	3.2	4.8	5.6	4.8	18.4	2.4	3.6	4	4.4	4.8	5.1
SLAM2025.09		1.6	2	1.6	1.6	1.2	2	2.4	1.6	1.2	4	2.4	4.4	2	2.8	2.4	1.6	2.2
MetaMath-Mistral 7B2025.09		1.2	2.4	2	2	2.4	3.6	4.4	2	1.2	2.4	10.4	2.8	2.8	1.6	1.6	1.6	2.8