Large Language Model Evaluation in Traditional Chinese Medicine for Stroke: Quantitative Benchmarking Study
Bibliographic record
Abstract
BACKGROUND: The application of large language models (LLMs) in medicine is rapidly advancing. However, evaluating LLM capabilities in specialized domains such as traditional Chinese medicine (TCM), which possesses a unique theoretical system and cognitive framework, remains a sizable challenge. OBJECTIVE: This study aimed to provide an empirical evaluation of different LLM types in the specialized domain of TCM stroke. METHODS: The Traditional Chinese Medicine-Stroke Evaluation Dataset (TCM-SED), a 203-question benchmark, was systematically constructed. The dataset includes 3 paradigms (short-answer questions, multiple-choice questions, and essay questions) and covers multiple knowledge dimensions, including diagnosis, pattern differentiation and treatment, herbal formulas, acupuncture, interpretation of classical texts, and patient communication. Gold standard answers were established through a multiexpert cross-validation and consensus process. The TCM-SED was subsequently used to comprehensively test 2 representative LLM models: GPT-4o (a leading international general-purpose model) and DeepSeek-R1 (a large model primarily trained on Chinese corpora). RESULTS: The test results revealed a differentiation in model capabilities across cognitive levels. In objective sections emphasizing precise knowledge recall, DeepSeek-R1 comprehensively outperformed GPT-4o, achieving an accuracy lead of more than 17% in the multiple-choice section (96/137, 70.1% vs 72/137, 52.6%, respectively). Conversely, in the essay section, which tested knowledge integration and complex reasoning, GPT-4o's performance notably surpassed that of DeepSeek-R1. For instance, in the interpretation of classical texts category, GPT-4o achieved a scoring rate of 90.5% (181/200), far exceeding DeepSeek-R1 (147/200, 73.5%). CONCLUSIONS: This empirical study demonstrates that Chinese-centric models have a substantial advantage in static knowledge tasks within the TCM domain, whereas leading general-purpose models exhibit stronger dynamic reasoning and content generation capabilities. The TCM-SED, developed as the benchmark for this study, serves as an effective quantitative tool for evaluating and selecting appropriate LLMs for TCM scenarios. It also offers a valuable data foundation and a new research direction for future model optimization and alignment.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".