Syndrome Differentiation on TCM-SD benchmark leaderboard · SOTA2 Research

Syndrome Differentiation on TCM-SD

82.97AccuracyMacBERT + dynamic H1 weights

Evaluation Results

Method
MacBERT + dynamic H1 weights2026.07	82.97	32.88	91.67
MacBERT + extracted entities2026.07	81.76	31.69	90.01
BERT + retrieved similar cases2026.07	81.38	31.45	90.26
MacBERT + fixed H1 weights2026.07	81.18	31.05	91.3
MacBERT + retrieved similar cases2026.07	81.14	31.17	90.52
BERT + extracted entities2026.07	80.95	31.18	90.12
Medical / TCM-domain BERT2026.07	80.72	29.05	90.88
BERT + fixed H1 weights2026.07	80.02	30.86	91.05
BERT + dynamic H1 weights2026.07	79.93	30.2	90.63
BERT-base Chinese2026.07	79.73	30.47	90.61
MacBERT2026.07	79.15	27.43	90.21
RoBERTa-wwm-ext2026.07	78.84	28.76	90
Qwen2.5 RAG + candidate syndrome list2026.07	78.19	30.37	90.16
Qwen2.5 + candidate syndrome list2026.07	74.26	28.64	87.55
Qwen2.5 few-shot2026.07	64.18	25.69	78.63
Qwen2.5 zero-shot2026.07	58.42	23.86	73.15
BiLSTM-Attention2026.07	48.51	25.38	66.03
H1 evidence only2026.07	32.7	21.72	53.61
TextCNN2026.07	31.47	20.16	50.12
Majority / frequency label2026.07	28.34	19.75	48.33
TF-IDF + SVM2026.07	27.93	20.07	48.15