TCM LLM · Datasets
TCM LLM datasets
Public datasets for training and evaluating TCM LLMs: instruction data, case QA, formulas, classical corpora, and benchmarks.
TCM-RobustSDT
TCM-RobustSDT: a robustness benchmark dataset for LLM clinical reasoning in TCM (Figshare).
HSQ-TD（健身气功指令微调数据集）
HSQ-TD: the first instruction-tuning dataset for health Qigong/wellness, with 57,843 instructions distilled from official textbooks and professional literature (ScienceDB).
中医体质量表计分代码
SAS code, SPSS syntax, and an Excel test sheet for scoring the Constitution in Chinese Medicine Questionnaire. Released January 2021, CC BY 4.0.
五部伤寒书知识图谱
A knowledge graph drawn from Shanghan Lun, Shanghan Leifang, Shanghan Xuanjie, Shanghan Lun Qianzhu, and Shanghan Jiushi Lun. It has 26,771 nodes and 8,272 relations in CSV files that can be loaded into Neo4j. Released on ScienceDB in 2024.
Tao-Shanghan-Corpus
A downloadable Shanghan Lun corpus that keeps clauses, commentaries, and textual variants apart. Version 1.0, released 11 August 2026, has 681 clause records, including 398 canonical clauses, 2,958 aligned commentaries, and 616 variants. CC BY 4.0.
中医世家
An online reader for TCM classics, herbs, formulas, and case records. The formula index has 25,425 entries, browsable by initial character or pinyin, and the herb index has 11,307 entries.
NPASS
A database of measured activity for natural products. This version lists 204,023 natural products, 48,940 source organisms, 8,764 targets, and 1,048,756 activity records, with values drawn from ChEMBL and the literature. Species records were also taken from TCM-ID, TCMID, TCMSP, and TM-MC. The 2026 update appeared in Nucleic Acids Research in November 2025.
TCM-Suite
A site that joins herb identification with network pharmacology. The Watson section currently lists 6,692 formulas, 7,322 herbs, 704,321 compounds, 19,319 proteins, and 15,437 diseases. Holmes identifies herbs with marker genes. Described in iMeta in 2022.
古代经典名方关键信息表
Official key-information tables for ancient classic formulas, not patient records. The notices cover 7 formulas in 2020, 25 including Zhuye Shigao Tang in 2023, and 43 including Xuanfu Daizhe Tang in 2025. The 2025 table gives source, processing, and gram doses. In Xuanfu Daizhe Tang, licorice is dry-fried (41.40 g) and pinellia is Qing Ban Xia (34.50 g). The decoction uses 2000 mL of water, reduces to 1200 mL, removes the marc, boils again to 600 mL, and is taken warm at 200 mL three times a day.
SYD_NK 内科证候测试集
366 TCM internal-medicine test items. Each item is a syndrome name plus a first-person symptom narrative. The card says the text was cleaned from an internal-medicine textbook and checked by TCM students. Apache-2.0.
治法–处方与药材剂量表
The repository readme describes 23 treatment methods, more than 400 subcategories, 4,174 common medicinal materials with doses and sources, and more than 900 classic formulas, with 21 feature categories. Files include Treatment.csv and a treatment–prescription workbook.
中药化学成分毒性实验库
Literature-derived in vivo and in vitro toxicity experiments for chemical constituents of traditional Chinese medicines, including Naoyanghua, Banmao, croton, Yangjinhua, Fuzi, and Qingfen. Fields cover route, species, indicators such as LD50, toxic dose, target organ, pathology, and PMID.
中药异构网络结构集
Heterogeneous biomedical graphs that include herb–target, herb–symptom, herb–herb, and symptom–symptom networks, together with protein–protein interaction networks.
MetaAcuPoint
900 synthetic RGB hand images from MetaHuman avatars in Unreal Engine, with bone-attached sockets for five hand acupoints used in Korean medicine.
TCMToxDB
A toxicity database that separates toxic effects of formulas, herbs, and ingredients, and gathers targets for specific toxic manifestations. Five herb–target and ingredient–target prediction algorithms are included. Site: https://www.sdu-idea.cn/TCMToxDB.
健康成人韩医诊察数据集
A traditional Korean medicine examination protocol, built from the literature and three rounds of expert review, with healthy-adult examination data collected over four years.
中医药中英术语数据集
A Chinese–English terminology dataset for traditional Chinese medicine, published in China Scientific Data.
针灸临床实践指南数据集
A dataset of acupuncture clinical practice guidelines, released to make the guidelines easier to use in clinic.
TCMCPD
Structured case-derived tables: 4,895 patients, 12,900 visits, 703 herbs, 11,879 visit–syndrome links, and 63,192 visit–herb rows. Dose units are mostly grams (61,386), plus 1,067 qian, 364 pieces, 240 tablets, 69 liang, 64 fen, and 2 millilitres. Rows also link symptoms, ingredients, and targets, with Neo4j files. Full copyrighted case texts are not redistributed. The describing manuscript is a Research Square preprint.
赵汉青门诊处方（2025）
De-identified 2025 internal-medicine outpatient prescriptions from Zhao Hanqing: 122 patients, 335 prescriptions, and 5,951 herb rows with gram doses. 75 are initial visits and 260 are follow-ups. Routes: 315 oral decoctions, 8 oral dissolved, 8 external washes, and 4 external uses. Dosing is twice daily in 333 prescriptions and three times daily in 2. Herb-name rows containing processing marks, which can overlap: stir-fried 412, honey-fried 341, bran 289, vinegar 164, ginger 154, wine 89, calcined 67, prepared 63, and processed Pinellia (Fa Ban Xia) 99. CC BY 4.0. The repository describes the clinic as Beijing Heniantang Hospital.
TCM4NSCLC
Computable real-world outpatient records from one senior TCM physician on non-small-cell lung cancer, for syndrome differentiation, prescription, and real-world effectiveness.
hwtcm-deepseek-r1-distill-data
Haiwei AI Team's DeepSeek-distilled TCM reasoning QA set built from web text. The authors state that it was not manually reviewed.
hwtcm-sft-v1
Haiwei AI Team's 7,096-example TCM instruction set. Text from websites, forums, and books was converted to QA and scored for expertise, safety, and fluency.
TCM-QM
Gas-phase DFT descriptors for 3,196 TCM-related molecules, with 16,918 deduplicated herb–molecule provenance links.
医宗金鉴知识三元组
25,571 knowledge triples extracted from the Golden Mirror of Medicine, covering medicinals, formulas, diseases, acupoints, and theory across 1,105 relation types.
TongueAtlas-4K
A 4,000-image clinical tongue benchmark with certified TCM expert labels for 22 appearance categories, released for MIRNet (AAAI 2026, to appear). The Zenodo record is restricted.
VTM-TDA Benchmark
A FAIR computational benchmark of Vietnamese traditional medicine digitized from Do Tat Loi's pharmacopoeia, with 724 harmonized botanical entities and 1,635 historical prescriptions.
VietMedKG
Knowledge graph and benchmark for traditional Vietnamese medicine.
TCMDI
Benchmark for predicting contraindicated interactions between traditional Chinese medicines and drugs.
六经辨证数据集
Traditional Chinese medicine dataset for six-meridian syndrome differentiation.
TreeProbe
Tibetan-medicine benchmark for measuring cultural bias in large language models.
四模型中文医学文献标引测试集
Key-information extraction and summary scores from four large models on 100 Chinese medical research articles.
病机思维链证候识别数据
Syndrome-manifestation cases with gold labels and model predictions, built under pathogenesis reasoning chain-of-thought supervision.
中医药知识问答实验数据
Traditional Chinese medicine question-answering set plus the knowledge graph, prompt-engineering code, and experiment outputs.
TCM-MS2Link
AI-ready dataset joining curated herb–compound links (TCM-MolLink) with MS/MS spectra.
病理标注舌象数据集
Tongue-image dataset with pathological annotations for AI-assisted TCM diagnosis.
中药扰动转录组图谱
Harmonized transcriptomic atlas of traditional Chinese medicine perturbations, assembled from public treatment experiments.
胆石症中医证候代谢组
Serum and urine metabolomics for two TCM syndromes in cholelithiasis: liver qi stagnation and liver-gallbladder damp-heat.
TCMP-300
Image dataset of traditional Chinese medicinal plants for plant recognition.
传统方剂—疾病关系标注语料
Annotated corpus of traditional formula–disease relations in biomedical articles. Formulas are herb combinations prepared under traditional medicine principles.
AcuSim
Synthetic dataset for localising cervicocranial acupuncture points, built because expert annotation is costly and limits detector training.
Baize TCM Botanical Origin Images
Medicinal-plant origin images for Baize Traditional Chinese Medicine, covering habitats, original plants, and medicinal parts, with Chinese and Latin identification based on the Chinese Pharmacopoeia, Flora of China, and the Engler system.
HerbPairIAM
Data and model artifacts for attributing adverse-reaction signals to components of multi-component Kampo medicines, with a zip archive and README on Zenodo.
CHM-315
Fine-grained Chinese herbal-medicine image set with 315 categories and 126,000 images, 400 images per category.
TCMEFO
Formal ontology for traditional Chinese medicine efficacy, providing a hierarchical computable framework over concepts, relations, and herbal-medicine instances.
TCM Medical Case Commentary Generation
Generation set for TCM medical-case commentary, with 510 standardized source cases and 2,550 training and test instances compiled from gynecological classics.
TCM Knowledge Q&A
Science Data Bank set of 6,157 question-answer pairs built from classic Chinese medicine literature.
TongueBench
Quality-controlled tongue-image set of 2,694 photographs with file-level metadata; the open file is TongueBench Dataset.zip.
Chinese Herbal Efficacy Relations
Structured herb–efficacy relations converted from classical TCM efficacy wording for knowledge-graph construction, formula-efficacy prediction, and herb–symptom association.
TCM Term Prompt-Language Responses
972 model responses for 18 TCM terms across six general-purpose models, three prompt languages (Chinese, English, and Malay), and three runs, with WHO 2007 and 2022 terminology references.
WHO TCM Terminology LLM Evaluation Set
Evaluation set for using large language models as judges of TCM terminology, containing 376 contrasts from the 2022 WHO International Standard Terminologies on Traditional Chinese Medicine.
TCM Disease-Pattern-Symptom-Herb KG v2
Version 2.0 semantic knowledge graph linking diseases, patterns, symptoms, and herbs, released as node and edge CSV files.
Cross-linguistic TCM Formula Names
Cross-linguistic table of TCM formula names in Chinese, Japanese Kampo, Korean Han medicine, and English, compiled from official formularies and clinical references in Taiwan.
TCM Humanities MCQ
Multiple-choice questions on medical humanities, ethics, and physician regulations, curated by the Traditional Chinese Medicine Language Model Team; the card marks the size as under 1,000.
tw-tcm
minyichen/tw-tcm card lists 8,145 training examples with question, answer, source, and think fields.
TCM_19W_DataSet-SFT (part)
The card describes a partial non-web TCM instruction-tuning set, released as TCM19W_Part1.json.
TongueProjectTCM Color Classification
TongueProjectTCM tongue-image set whose card lists images and integer labels, with 807 training and 121 validation examples.
Chinese Drug Name Overreliance Set
Experiment data for the paper on whether LLMs over-rely on names to identify ingredients of Chinese drugs, linked from the dataset card to arXiv:2504.03786.
CMMLU Traditional Chinese Medicine
CMMLU Traditional Chinese Medicine test split; the card shows a 185-example load and the file traditional_chinese_medicine.json.
JTB Taiwan TCM Exam Questions
JTB Tech compilation of Taiwan Ministry of Examination TCM questions for ROC years 90–113, covering TCM internal medicine, basic medicine, and clinical medicine.
Taiwan TCM Examination MCQ
Past national TCM exam questions from Taiwan's Ministry of Examination; the README gives ROC years 96–111 excluding 110, and the repository file is basic theory of traditional Chinese medicine.csv.
2023 Pharmacist Licensure Examination TCM Track
TCM Pharmacy track items from the 2023 Chinese National Pharmacist Licensure Examination; the card says they were collected online and points to FreedomIntelligence/HuatuoGPT-II.
TCMLE
Chinese TCM practitioner licensing question bank with 7,956 items: 2,700 assistant-practitioner and 5,256 practitioner questions, including past papers and mocks.
Famous Old TCM Doctors Cases
Outpatient cases from Contemporary Famous Old Chinese Medicine Doctors' Typical Cases Collection; the card says each case includes the illness description, syndrome differentiation, and prescriptions.
tc_medical_base
Clause-level corpus of five classical TCM texts totaling 3,569 clauses with sentence-BERT embeddings: Song-edition Shanghan Lun, Guilin Shanghan Zabing Lun, Jingui Yuhan Jing, Jingui Yaolue, and Fu Qingzhu Nüke.
TriCM-DF Acupuncture Corpus
Expert-verified instruction corpus for TCM models, instantiated for acupuncture across consensus, experiential, and perception tables; the repository releases text, not the image or video files.
Gancao-TCM-Instruct
Public Gancao-TCM-Instruct release with 25,000 cases and 125,000 samples covering disease diagnosis, syndrome diagnosis, treatment selection, base-formula selection, and herb selection.
AcuKG
Knowledge graph of medical acupuncture assembled from guidelines, literature, trials, and ontology mapping.
KampoBench
Physician-reviewed Kampo consultation benchmark with 74 scenarios and 310 dialogue-level criteria.
TCMClinicalReason-Bench
2,000 multicenter records for reasoning from pathogenesis to prescription.
TCMQA
38,279 TCM licensing questions with 15,151 responses from 101 licensed practitioners.
辨证论治基准数据集
Scientific Data benchmark for syndrome differentiation and treatment with large language models.
TongueDx
Public remote-tongue dataset: 5,109 images from 4,650 subjects, eight binary attributes (pale, red tip/side, red spots, ecchymosis, cracks, tooth marks, thick fur, yellow fur), plus age and sex. The repository is downloadable. This is not the later TongueDx2 ablation paper.
内蒙古自治区民族药目录
Public ethnic-medicine catalogue PDF from the Inner Mongolia Healthcare Security Administration, mainly finished Mongolian formulas. The government filename is dated 25 December 2023. The extracted text is titled the Inner Mongolia ethnic-medicine catalogue and notes that listed drugs are managed as class B. It is a reimbursement list, not a research compound database.
傣药基原植物数据库
Dai medicinal-plant source database on the Chinese Union of Botanical Gardens platform, from Xishuangbanna Tropical Botanical Garden, CAS. The home page is copyrighted 2018 and describes Dai medicine’s four towers, five aggregates, and antidote theory. The species-list URL returns a redirect notice, and no anonymously browsable species catalogue was confirmed.
Sowa Rigpa Materia Medica Dataset
Open Data in Brief table of Latin identifications for commonly used Sowa Rigpa materia medica. The paper says about 700 commonly used items were selected from the Four Medical Treatises and Gawé Dorjé’s classification, with Tibetan script, Wylie transliteration, and Latin names; the supplement is published with the article. The authors are at the School of Tibetan Medicine, Gansu University of Chinese Medicine, the University of Wisconsin-Madison, and the College of Pharmacy at Qinghai Nationalities University. The data-source location is the ethnic-medicine center at Chengdu University of Traditional Chinese Medicine.
藏药材资源数据库
Tibetan materia-medica list published by the Key Laboratory of Tibetan Medicine Research at the CAS Northwest Institute of Plateau Biology. The page is a table of Chinese name, Latin name, drug name, and part used, and it states that there are 15 pages. Entry paths are dated March–April 2022. It is a crude-drug list, not a compound–target database.
KMCRIC
Data portal of the Korean Medicine Convergence Research Information Center at Kyung Hee University. The data-center page lists an evidence-based Korean-medicine literature set of systematic reviews, RCTs, and mechanism studies published since 2010; an herb-interaction set based on Natural Standard; acupoint videos made from the 2008 WHO Western Pacific standard locations; and a crude-drug update built with the Kyung Hee and Chungbuk pharmacy faculties. The director’s biography says the directorship began in 2013.
韩医药素材银行药材目录 (KMMB)
Korean Medicine Material Bank at the National Institute for Korean Medicine Development. On 1 January 2016 it was formed by merging the natural-substance bank and the herbal biomaterial bank. The herb page lists 691 items, with name, synonym, scientific name, botanical source, and reacting enzymes.
KTKP Traditional Medicine
Korean-medicine slice of the Korean Traditional Knowledge Portal, run for the Ministry of Intellectual Property. The English introduction says the database was compiled from traditional Korean medicine literature in 2005–2007 and opened for search in December 2007. The same page currently counts 5,500 herbs, 7,103 prescriptions, and 12,500 diseases, linked to effects, indications, formulas, and patent classes.
ETHMEDmmm
Ethnomedicine specimen database of the Museum of Materia Medica at the University of Toyama. The home-page table lists 461 kinds and 8,942 Japanese–Chinese crude-drug specimens, opened in Japanese in April 2000. The same site also has an English crude-drug list and separate Unani, Thai, Ayurvedic, and Indonesian collections. The latest note is a version-7 rebuild on 30 March 2018. The site returned 503 on this check; the institute portal still lists the address.
EKAT
Evidence Reports of Kampo Treatment, a structured-abstract database from the Japan Society for Oriental Medicine EBM committee. The site says it has been free and updated since 2007. It covers RCTs and meta-analyses of Kampo products still on the market, limited to reports since 1986 that used the same quality as marketed products; decoctions are excluded because product quality is unknown. Japanese and English editions are provided, and the Association of Korean Medicine has a Korean translation.
药用植物综合信息数据库 (MPDB)
Comprehensive medicinal-plant database of the NIBIOHN Research Center for Medicinal Plant Resources. The home page describes medicinal plants, crude drugs, constituents and genes, cultivation, and internal morphology, plus measured data for representative crude drugs in Japanese circulation. The footer copyright is 2012–2025, funded by a MHLW research grant. Separate Kampo-formula and Japanese Pharmacopoeia crude-drug lists are online.
KampoDB
Kampo mechanism database maintained by the Institute of Natural Medicine, University of Toyama. It links formulas, crude drugs, natural compounds, and target proteins, including known interactions and docking predictions. The about page currently lists 298 Kampo formulas, 180 crude drugs, 3,002 natural compounds, 62,906 proteins or genes, and 3,063,505 docking results. The 2018 Scientific Reports paper is the citation; the site news records a renewal on 15 July 2021.
TradMPD
Traditional Medical and Pharmaceutical Database at the University of Toyama Institute of Natural Medicine. A MEXT regional innovation cluster grant supported it in 2008–2012. It covers crude-drug identity, properties, Kampo composition and indications, and more than 10,000 Kampo papers. The Wakanyaku Library holds nearly 80 compounds plus chemical and bioassay data for hot-water extracts of 80 crude drugs and 80 Kampo extracts, with bioassay downloads from 2024. The database site returned 503 on this check; the institute portal still lists the address.
中药方剂图像数据库
Formula image library from the School of Chinese Medicine, Hong Kong Baptist University, openly browsing composition, actions, indications, and images. The browse page shows about 182 formulas, and the public counter starts in November 2017. It is an online reference, not a bulk formula-table download.
中药材图像数据库
Chinese medicinal-material image library from the School of Chinese Medicine, Hong Kong Baptist University. The library page states more than 420 commonly used crude drugs, with source, origin, morphology, properties, indications, and high-resolution images, searchable by Chinese, Latin, and English names. It has been online since 2013 and is openly browsable, not a bulk download.
TCMSSD
Syndrome-standardization database reported by authors at Zhejiang Chinese Medical University. The paper reports 624 syndromes, 133,518 prescriptions, 8,073 diseases, 8,259 herbs, 43,413 ingredients, and 17,602 targets. The site opens and its front-end script contains syndrome and TCMSSD content; without JavaScript the page is empty. No full-database download entry was found.
TCMPG 2.0
Traditional Chinese medicine plant-genome database maintained by Chengdu University of Traditional Chinese Medicine and collaborators. The current homepage lists 274 plants, 324 genomes, 414 herbs, and 13,809 ingredients, with BLAST, JBrowse, and other tools. A 2024 paper describes the 2.0 upgrade. The site is open for browsing and analysis; a full packaged dump was not verified.
本草典开放数据
Bilingual TCM seed data from the Bencao Dian editorial group, released under CC BY-SA 4.0 as directly downloadable JSON. Version 1, generated on 17 April 2026, lists 365 herb records, 112 formulas, 84 patterns, and 208 acupoints, plus meridians, tongue and pulse entries, and a relation graph. A coverage note on the same page says 335 of about 530 standard materia medica items are included. This is an editorial vocabulary, not a national standard.
SoFDA
Syndrome-ontology platform built by the Institute of Chinese Materia Medica, CACMS, and collaborators. The iMeta 2023 paper describes 319 curated syndromes, 8,045 diseases, and 1,359 formulas, linked to genes, diseases, and formulas. The paper page checks out; the front page currently shows no readable syndrome content. It shares the tcmip.cn host with ETCM but is a separate syndrome system. No anonymous full-database mirror was found.
GB/T 16751.3-2023
Current national standard Clinic terminology of traditional Chinese medical diagnosis and treatment — Part 3: Therapeutic methods, issued and effective on 17 March 2023. The official full-text disclosure system offers online preview and download. This is the therapeutic-method standard page, not a separate vocabulary dump.
GB/T 16751.2-2021
Current national standard Clinic terminology of traditional Chinese medical diagnosis and treatment — Part 2: Syndromes/patterns, issued and effective on 26 November 2021. The official full-text disclosure system offers online preview and download. This is the syndrome-terminology standard page, not a separate vocabulary dump.
GB/T 16751.1-2023
Current national standard Clinic terminology of traditional Chinese medical diagnosis and treatment — Part 1: Diseases, issued and effective on 17 March 2023, under the National Administration of Traditional Chinese Medicine. The official full-text disclosure system offers online preview and download. Syndrome and therapeutic-method parts have their own full-text pages.
GB/T 15657-2021
Current Chinese national standard Classification and codes of diseases and patterns of traditional Chinese medicine, issued and effective on 11 October 2021, replacing GB/T 15657-1995, under the National Administration of Traditional Chinese Medicine. The official full-text disclosure system offers online preview and download. This is the standard page, not a standalone machine-readable code package.
GB/T 12346-2021
Current Chinese national standard Nomenclature and location of meridian points, issued and effective on 26 November 2021, under the National Administration of Traditional Chinese Medicine. The official full-text disclosure system offers online preview and download. This is the standard page, not a separately packaged acupoint table.
WHO西太区标准穴位定位
WHO Regional Office for the Western Pacific standard for acupuncture point locations, published in 2008 and reprinted in 2009. The IRIS record states the surface-location method and 361 points for teaching, research, clinical service, publication, and academic exchange, and provides the English PDF. Experts from China, Japan, and Korea joined the location consensus. The record page is public.
ICD-11传统医学章节
Official WHO explanation of the ICD-11 supplementary chapter Traditional Medicine Conditions. The chapter is for optional dual coding of morbidity and is not used for mortality. Module I covers traditional-medicine conditions that originated in ancient China and are now used in China, Japan, Korea, and elsewhere. This page explains the chapter; the codes themselves are read in the ICD-11 browser. The page is public.
国家中医药古籍数字图书馆
Ancient-book and traditional-knowledge portal started in 2022 by the Institute of Information on Traditional Chinese Medicine, CACMS. It has halls for classical texts, Republican-era documents, and bibliographic catalogs, and links out to reading systems including Guoyi Diancang. The portal opens; full text is reached through the sub-libraries and is not a public bulk download.
中华医典
Commercial sales site (Feixiong Technology) for the Zhonghua Yidian electronic collection. The site states 1,156 classical titles and about 450 million characters, a 2000 national electronic-publication nomination, and Qiu Peiran as editorial reviewer and title writer. Edition 6 is a web version used after purchase and registration, not a public download.
国医典藏
Image-and-text library of historical Chinese medical books from the Institute of Information on Traditional Chinese Medicine and its library, CACMS. The current entry is the version-4 site. Public descriptions put the holdings at roughly 2,500 to 3,000 titles from pre-Qin through the late Qing and Republican periods, with basic, advanced, and controlled-vocabulary search and image-text viewing. Access is by campus network or registration, not a public bulk download.
中医药知识服务平台
Knowledge-service homepage maintained by the Institute of Information on Traditional Chinese Medicine, CACMS, for search, knowledge graphs, and disease profiles. A related about page from the same institute says that, starting in 2019, it built a professional TCM knowledge service for the China Knowledge Centre for Engineering Sciences and Technology, bringing together diseases, formulas, herbs, and literature. The homepage opens; campus trial notes require registration on an institutional network, and no public bulk download was found.
古今医案云平台
Case search, structuring, and mining platform of the Institute of Information on Traditional Chinese Medicine, China Academy of Chinese Medical Sciences. An institute news item on 2026-09-20 states that the current resource holds more than 1.1 million structured historical and modern cases, plus 19.63 million analysis records, with about 53,000 registered users. Registration or institutional login is required, and no public bulk download was found. Some campus trial notes still say about 500,000 cases.
OASIS (KIOM)
KIOM traditional-medicine literature portal for Korean-medicine papers and herbal resources. The Korean-medicine paper search went online in June 2007 and was rebuilt as a traditional-medicine portal in July 2009.
Korean Medicine Embedding Dataset
Query–positive–negatives (~113k pairs) built from Korean-medicine terms and an ontology, for embedding fine-tunes such as BGE-M3.
KNApSAcK KAMPO
NAIST Kampo public database (~1,581 formulas, 278 crude drugs), downloadable from the NBDC life-science archive.
webMedQA
Early Chinese non-factoid medical QA from health-consult sites (~63k questions, one positive and four negative answers each).
IMCS-21
About 4,116 pediatric online consults annotated for entities, intents, symptoms, and reports; later wired into four CBLUE dialogue tasks.
CBLUE
Chinese biomedical NLU benchmark (NER, relations, diagnosis normalization, classification); the source-task suite behind PromptCBLUE, with a Tianchi submission portal.
CMeKG
Chinese medical knowledge graph of diseases, drugs, and symptoms; main source for BenCao/HuaTuo and ChatGLM-Med instruction data. Official portal is unstable; verify via the tools repo.
MedDialog
Large doctor–patient dialogue corpus (about 1.1M Chinese encounters) used for multi-turn Chinese medical fine-tuning.
cMedQA2
Chinese community medical QA (~108k questions / 200k answers), a common source for BianQue-style SFT mixtures.
cMedQA
Chinese community medical QA-matching set (repo table ~54k questions / 102k answers; non-commercial research). Paper DOI matches the README; see cMedQA2 for the later release.
ChiMed (Qilin)
Qilin-Med's ~3GB Chinese medical corpus (CPT/SFT/DPO); not the same resource as the ChiMed 2.0 pretraining set.
DISC-Med-SFT
Fudan DISC medical-dialogue SFT set (~470k examples from KG triples and reconstructed consults; no preference data).
PromptCBLUE
CBLUE's 16 Chinese medical NLP tasks rewritten as generative instructions; an early unified Chinese medical LLM leaderboard (CCKS 2023).
Huatuo-26M
Largest open Chinese medical QA resource (~26M pairs from encyclopedias, KGs, and consults); Huatuo-Lite is the usual SFT/RAG subset.
CMExam
Chinese medical licensing-exam set (~68k annotated items) used as a knowledge-recall baseline by Chinese medical and TCM LLMs.
CMB
FreedomIntelligence comprehensive Chinese medical benchmark (CMB-Exam ~280k items plus CMB-Clin cases); the most common non-TCM comparison board in TCM LLM papers.
TM-MC
KIOM's literature-derived database of Northeast Asian medicinal materials and compounds. The 2015 release covers about 536 materials, and the 2024 version 2.0 expands to about 34,000 compounds. The old address says the data and services have moved to tm-mc.kr.
CVDHD
3D structures of 35,230 natural products from cardiovascular herbs, docked to 2,395 target proteins, with related diseases, pathways, and clinical biomarkers. The original Peking University site timed out on this check.
TCMGeneDIT
Text-mined links among TCM, genes, diseases, effects, and ingredients, joined to pathways and protein interactions. tcm.lifescience.ntu.edu.tw does not resolve on this check.
TCM-Mesh
Network-pharmacology resource with 6,235 herbs, 383,840 compounds, 14,298 genes, 6,204 diseases, 163,221 side-effect records, and 71 toxicity records. The old site returned 403 on this check.
TCMAnalyzer
Sun Yat-sen University toolkit for scaffold search of bioactive ingredients and for networks of molecules, targets, pathways, and diseases. rcdd.org.cn timed out on this check.
SuperTCM
Charité resource of 6,516 TCM drugs, 5,372 botanical species, 55,772 active ingredients, 543 targets, 254 KEGG pathways, and 8,634 diseases. tcm.charite.de does not resolve on this check.
CMAUP
BIDD landscape of multi-target activities, pathways, and diseases for useful plants including TCM herbs; 2024 update, downloadable from the official site.
DCABM-TCM
Literature-mined blood constituents and metabolites of TCM prescriptions and herbs, with experimental detection conditions (~1,816 structured absorbed constituents).
ITCM
Integrated formula/herb/ingredient/target platform plus 1,488 pharmacotranscriptomic profiles for 496 TCM ingredients (expression data also on Synapse).
TCMBank
Large downloadable herb–ingredient–target–disease resource with literature-mining updates after manual checks.
YaTCM
1,813 prescriptions, 6,220 herbs, and 47,696 natural products, with target and pathway tools. The Nankai site returned 403 on this check.
CEMTDD
An herbal–compound–target–disease resource centered on Uygur and Kazakh medicines. The paper reports about 621 herbs.
TCMIO
Immuno-oncology TCM database of prescriptions, herbs, ingredients, targets, and pathways, with downloads and a REST API.
LTM-TCM
Links 14 TCM databases with clinical records and classical books across symptoms, prescriptions, plants, ingredients, and targets: 1,928 symptoms, 48,126 prescriptions, 9,122 plants, 34,967 ingredients, and 13,109 targets. cloud.tasly.com does not resolve on this check.
HIT 2.0
Manually curated herbal-ingredient–target activity pairs (~1,237 ingredients / 2,208 targets, 2000–2020 literature). Portal opens; the analysis backend port is currently down (marked site issue).
TCM Database@Taiwan
More than 20,000 pure compounds isolated from 453 TCM ingredients, each in 2D and 3D, for virtual screening. tcm.cmu.edu.tw timed out on this check.
BATMAN-TCM 2.0
Known and predicted TCM ingredient–target protein interactions, with greatly expanded TTI coverage and target-to-ingredient search.
SymMap 2.0
Maps among herbs, TCM symptoms, modern symptoms, ingredients, targets, and diseases. The site is now version 2.0, updated with the 2020 Chinese Pharmacopoeia, and the relationship tables can be downloaded. The footer is dated 2022.
HERB 2.0
Evidence-centered TCM resource integrating clinical trials, meta-analyses, high-throughput experiments, literature, and a knowledge graph.
ETCM 2.0
Encyclopedia of TCM formulas, patent drugs, materia medica, and ingredients with target prediction and multi-scale networks; v1 site remains online.
TCM-ID
NUS BIDD formula–herb–ingredient–target resource covering pharmacopoeia, classical, and CFDA-approved prescriptions; not the same database as TCMID 2.0.
TCMID 2.0
East China Normal University's database of formulas, herbs, ingredients, and targets. The data and the paper are on Zenodo and in Nucleic Acids Research. It is a different database from NUS TCM-ID. megabionet.org/tcmid timed out on this check.
TCMSP
A pharmacology site for herb ingredients, targets, diseases, and ADME numbers. The site now keeps version 2.3 beside the newer 9.0.1, and the relationship tables can be downloaded.
TCM-QG
Question-generation data from TCM texts: 3,500 documents, with one to four human-written question-answer pairs each. The license is CC BY-NC 2.0.
TCM-PD
Yao et al. TKDE 2018 prescription topic-model set (98,334 raw / 33,765 processed symptom–herb ID pairs). CKCEST copyright, research use only. PresRecST's prescript_1195.csv is a reproduction table.
TCM-Lung
Pulmonary-disease cases from FAH-HUCM (14,948 processed; 4,484 encoded public rows of symptom/syndrome/method/prescription IDs). Full names on request. Not the same resource as TCMNSCLC.
TCM-SD
First large public TCM syndrome-differentiation text benchmark (54,152 real records, 148 syndromes, CC BY-NC-SA 4.0). Full set is in the repo folder TCM_SD_with_knowledge; Tianchi id 139034.
TCM-NER
Named-entity data from Chinese medicine package inserts, for a medication knowledge graph. The license is CC BY-NC 2.0.
TCM_KG
A TCM knowledge graph built with Neo4j community edition. The sample query starts from the symptom node for constipation and returns its neighbors.
TCM-MKG
Open multidimensional TCM knowledge graph of terminology, herbal pieces, and medicinal properties. The companion paper uses interpretable graph neural networks on 6,080 formulas, with a check on 215 COVID-19 formulas.
LingShu（灵枢知识图谱）
Symptom-centric contextual KG bridging TCM and biomedicine (~17.33M entities, ~39.47M relations including triples and contextual quadruples), with a portal for visualization, reasoning, and evidence-grounded QA.
OpenTCM-KG
OpenTCM gynecology classics KG (~48k entities / ~152k relations).
TCMNSCLC
Real-world NSCLC TCM reasoning dataset with fully annotated cases (pattern differentiation / treatment method / decoction / patent medicine).
ChP-TCM
KnowledgeQA and PrescriptionWriting instructions built from Chinese Pharmacopoeia Vol. I.
Traditional-Chinese-Medicine-Dataset-SFT
High-quality TCM supervised fine-tuning dataset.
TCMChat-dataset-600k
TCMChat herbal QA and recommendation instruction data (~600k).
TCM-Instruction-Tuning-ShizhenGPT
ShizhenGPT multimodal SFT data (text/vision/speech/ECG etc.; ~311k items total per paper Table 3).
ShenNong_TCM_Dataset
ShenNong TCM instruction dataset.
MedChatZH
MedChatZH TCM consultation dataset.
ChatMed_Consult_Dataset
Chinese online-consultation set of 110,113 pairs. The questions come from consultation sites and the answers were written by GPT-3.5.
CMtMedQA
ZhongJing real multi-turn doctor–patient dialogues (~70k).
Baize-TCM-Corpus-V3
~157k TCM QA items covering theory, herbs, formulas, diagnosis, acupuncture, and clinic.
neijing-sft-v1.2
~2,009 Neijing-related instruction samples for Xinghe, with thinking/output fields.
TCM-Text-Exams
Recent TCM licensure / graduate-exam text benchmark.
Medical-LLMs-Chinese-Exam
Chinese medical exam evaluation for medical LLMs.
ZhongJing-OMNI
ZhongJing-OMNI multimodal TCM eval (including tongue).
TCMEval-SDT
TCMEval-SDT: a benchmark of 300 syndrome-diagnosis cases (web, classical texts, hospital records) for evaluating TCM syndrome-differentiation reasoning, with FAIR metadata (Sci. Data 2025).
TCMBench
TCMBench: a comprehensive benchmark for evaluating LLMs in traditional Chinese medicine (arXiv 2024).
TCM-Vision-Benchmark
TCM vision benchmark (herb recognition / inspection, ~7k items).
TCM-Tongue
6,719 standardized tongue images with 20-class multi-label pathology annotations and detection baselines.
TCM-Ladder
TCM-Ladder: a multimodal QA benchmark for comprehensively evaluating TCM multimodal LLMs on real-world tasks (arXiv 2025).
TCM-Eval
Dynamic, extensible TCM evaluation platform.
TCM-BEST4SDT
Case benchmark for syndrome differentiation and treatment.
TCM-5CEval
Five-dimension deep TCM evaluation suite.
TCM-3CEval
Three-axis eval: core knowledge, classics, clinical decisions.
MTCMB
MTCMB dataset: a multi-task TCM benchmark covering knowledge, reasoning and safety, 12 subsets with ~7,100 samples (arXiv 2025).
HWTCMBench
HWTCMBench TCM capability evaluation set.
TCMEval-PA
328 multiple-choice items on prescription normative quality and safety auditing.
TCM-AQA61
Synchronized first- and third-person videos of acupuncture and Tuina, 61 subjects each. Two experienced TCM physical therapists assessed every subject independently, then reconciled disagreements. Paired with the CME-AQA cross-view assessment framework.
LingLan
LingLan large multi-task TCM evaluation benchmark (2026).
ChiMed 2.0
Upgraded Chinese medical pretraining dataset covering TCM corpora for LLM pretraining.
classical-tcm-canon
Full-text digitizations of the TCM canon: Neijing, Nanjing, Shanghan Lun, Jingui Yaolue and warm-disease classics.
Traditional-Chinese-Medicine-Dataset-Pretrain
High-quality TCM pretraining dataset from non-Internet sources (~1GB; clinical cases, classics, encyclopedia), 99% simplified Chinese.
TCM-Pretrain-Data-ShizhenGPT
ShizhenGPT pretraining corpus. The paper reports 15B+ tokens — Stage 1 text 11.92B including 6.3B TCM, and Stage 2 multimodal data of 3.8B tokens in the main text and 3.84B in the appendix.
TCM-Ancient-Books
A corpus of nearly 700 TCM ancient-book texts.
awesome_Chinese_medical_NLP
Curated list of Chinese medical NLP resources: terminologies, corpora, word vectors, pretrained models, KGs, NER and QA (incl. CBLUE).
CPM中成药数据集
Living large-scale public Chinese patent medicine data accompanying RAG-CPMF.