Model Routing on Composite QA Suite (MathQA, LogiQA, MedQA, PIQA, TruthQA, MMLU, GSM8k, GPQA, ASDiv, SoQA) benchmark leaderboard · SOTA2 Research

Model Routing on Composite QA Suite (MathQA, LogiQA, MedQA, PIQA, TruthQA, MMLU, GSM8k, GPQA, ASDiv, SoQA)

66.2Overall AccuracyKMeans

Evaluation Results

Method	Links
KMeans2026.01		66.2	56.12	49.02	74.86	87.31	52.7	85.14	89.29	24.8	60.1	37.65
LOCUS2026.01		64.7	52.74	49.02	72.6	85.07	41.89	81.83	89.29	29.2	64.65	31.48
KMeans2026.01		64.47	57.81	52.94	75.42	87.31	47.3	81.32	89.29	23.6	57.58	37.65
IRT-NET2026.01		63.37	57.81	52.94	71.19	79.85	50	82.17	88.39	27.6	43.94	30.25
LOCUS2026.01		62.97	49.79	43.14	57.91	86.57	52.7	83.7	85.71	27.4	60.61	30.86
kNN2026.01		62.9	57.81	45.1	72.6	86.57	55.41	82.09	89.29	21.6	39.39	37.04
LOCUS2026.01		61.9	39.66	41.18	55.65	85.82	33.78	86.25	75.89	25.2	65.66	29.63
EmbedLLM+2026.01		61.57	57.81	45.1	73.45	81.34	41.89	78.27	89.29	29	33.84	32.72
kNN2026.01		61.37	57.38	45.1	68.64	85.82	48.65	80.48	89.29	26.8	29.8	29.01
KMeans2026.01		61.33	58.65	49.02	62.99	86.57	52.7	84.72	89.29	25.8	10.1	31.48
kNN2026.01		60.5	59.92	47.06	75.42	85.07	41.89	78.27	89.29	29	10.1	30.86
IRT-NET2026.01		60.17	58.65	49.02	72.6	80.6	44.59	77.25	83.04	28.4	23.23	32.1
EmbedLLM+2026.01		59.73	57.38	45.1	71.75	79.85	32.43	78.69	89.29	26.2	21.72	29.01
EmbedLLM2026.01		59.6	58.65	50.98	72.88	79.1	47.3	76.83	89.29	23.8	25.25	30.86
IRT-NET2026.01		59.57	59.49	39.22	72.6	81.34	43.24	76.91	90.18	28.6	14.65	30.25
EmbedLLM2026.01		59.47	59.49	45.1	74.29	77.61	43.24	76.83	89.29	28.8	12.12	29.63
EmbedLLM+2026.01		59.17	49.37	35.29	62.99	62.69	51.35	77.08	89.29	24.2	54.55	35.8
EmbedLLM2026.01		58.8	58.23	41.18	72.03	79.85	35.14	75.55	88.39	28.2	19.19	30.25