Medical Search Relevance Assessment on LocalQSMed v1.0 benchmark leaderboard · SOTA2 Research

Medical Search Relevance Assessment on LocalQSMed v1.0

97.6HR PrecisionOriginal System

Evaluation Results

Method	Links
Original System2025.12		97.6	70.31	81.74	7.56	59.84	13.43	7.53	22.45	11.28	69.28
Qwen3-32B2025.12		97.23	93.26	95.2	24	30.41	26.84	35.65	68.91	46.99	89.56
DeepSeek-R1-Distill-Qwen-14B2025.12		96.93	96.58	96.58	34.92	30.41	32.51	45.69	44.92	45.3	92.32
Qwen3-14B2025.12		96.47	95.94	96.21	28.86	19.82	23.5	38.1	67.23	48.63	91.48
Qwen3-8B2025.12		95.53	97.12	96.32	31.96	14.29	19.75	36.47	52.1	42.91	91.91
Qwen3-4B2025.12		95.41	97.41	96.4	26.02	14.75	18.82	44.8	47.06	45.9	92.06
DeepSeek-R1-Distill-Qwen-7B2025.12		95.13	21.73	35.37	16.22	3.57	5.85	30	3.03	5.5	20.4
Qwen3-0.6B2025.12		93.07	82.27	87.34	8.77	2.46	3.85	13.33	3.57	5.63	76.29
Llama-3.3-70B-Instruct2025.12		92.96	94.79	93.86	6.25	4.72	5.38	7.84	5.56	6.5	87.87
DeepSeek-R1-Distill-Llama-70B2025.12		92.93	93.02	92.97	8.33	5.41	6.56	5.5	7.5	6.35	86.26
Llama-3.1-8B2025.12		91.51	94.53	93	15.91	7.69	10.37	0	0	0	86.75
Llama-3.2-3B2025.12		91.43	93.24	92.32	2.17	1.16	1.52	4.17	2.17	2.86	85.65
Llama-3.2-1B2025.12		91.16	89.68	90.41	2.7	2.35	2.52	0	0	0	82.26