Haoxuan Li

Tsinghua Shenzhen International Graduate School

Unclaimed · AI-structured profileOpenAlex

AI for Traditional Chinese Medicine — LLM prescription generation, multimodal diagnosis, and knowledge graph systems

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-08) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Output metrics are not yet available: the data contains no publication counts or venue records beyond the three cataloged preprints. This axis would measure annual publication volume and venue tier once a fuller bibliographic record is linked.

Impact—

Impact cannot be assessed yet: no citation counts, h-index, or usage signals are present in the provided data. This axis would measure how often the researcher's work is cited or adopted, and requires a linked citation record.

Focus0/100

The focus score is 0 because the record shows no concentration: the three cataloged papers split evenly across medical AI/clinical NLP (TCM informatics), natural language processing, and database systems/knowledge graph data management, each field at 0.33. The work is unified by TCM applications rather than by method or venue, so the profile reads as application-driven breadth across AI and data systems rather than a single focused line.

Collaboration46/100

The collaboration score is 41, based on 7 distinct coauthors across the three cataloged papers, including Zheng Chen, Peiwu Qin, Zhicheng Du, Yingshan Liang, Yuzhu Li, Zhongde Zhang, and Lianshun Jin. On every paper Haoxuan Li is a middle-listed, non-corresponding author, suggesting contribution within a small stable team rather than leadership of an independent group.

Activity—

Activity data is insufficient: the provided snapshot covers only 2026 preprints and includes no yearly publication counts or multi-year trajectory. This axis would measure research cadence over time — papers per year and sustained output — and cannot be scored from the current evidence.

Reproducibility58/100

Reproducibility scores 58, a mean across 3 papers with mixed readiness: TCMaster is rated high, the TCM prescription-generation paper medium, and LingLan low. Evolve's verdicts note that reported gains — such as a claimed 103.5% relative accuracy gain (62.72%) and a 20-point QA win on a self-built benchmark — rest on inconsistent metric reporting or self-constructed evaluations, tempering confidence.

Research summary

Haoxuan Li is an early-stage AI researcher whose visible work centers on artificial intelligence for Traditional Chinese Medicine (TCM), spanning three complementary directions: LLM-based prescription generation, multimodal diagnosis modeling, and knowledge-graph data management. All three papers currently cataloged by Evolve are 2026 arXiv preprints on which Li is a middle-listed, non-corresponding author, working alongside seven distinct coauthors including Zheng Chen, Peiwu Qin, and Zhicheng Du. The first builds a structured-reasoning pipeline for TCM prescription generation using chain-of-thought distillation and DPO alignment with an emphasis on medication safety; Evolve notes its headline 7B-versus-GPT-5 comparison rests on inconsistent PQS reporting and narrow margins. The second, LingLan, is a multimodal TCM diagnosis LLM trained with supervised fine-tuning and LoRA for tongue and pulse diagnosis, reporting a 103.5% relative accuracy gain (62.72%); the third, TCMaster, contributes confidence-aware querying and workload-guided physical design over a 221K-node multi-source TCM knowledge graph. Evolve rates reproducibility readiness across these papers as one high, one medium, and one low, giving a mean reproducibility score of 58, while the field distribution splits evenly (0.33 each) across medical AI/clinical NLP, NLP, and database systems — breadth driven by the shared TCM application domain rather than by methodological focus.

Collaboration pattern

All three cataloged papers list Haoxuan Li as a middle author without corresponding-author status, within a compact circle of 7 distinct coauthors — Zheng Chen, Peiwu Qin, Zhicheng Du, Yingshan Liang, Yuzhu Li, Zhongde Zhang, and Lianshun Jin. The collaboration score of 41 reflects this small, stable team pattern; co-authorship spans all three TCM-AI and knowledge-graph papers, suggesting one shared research group rather than multiple independent networks.

Research keywords

TCM prescription generationlarge language modelschain-of-thought distillationDPO alignmentmedication safetymultimodal diagnosis LLMtongue and pulse diagnosissupervised fine-tuning and LoRAknowledge graphconfidence-aware queryingworkload-guided physical designCypher rewriting and path ranking

Highlights

TCMaster: confidence-aware querying over a 221K-node TCM knowledge graph
Workload-guided physical design plus Cypher rewriting and path ranking deliver honest, modest gains on multi-source TCM knowledge graphs; Evolve rates reproducibility readiness high, though the 20-point QA win rests on a self-built benchmark.
Structured reasoning pipeline for safe TCM prescription generation
Combines chain-of-thought distillation and DPO alignment around syndrome differentiation, synergy, and medication safety; Evolve finds the pipeline shows real effect, but the headline 7B-surpasses-GPT-5 claim relies on inconsistent PQS reporting and narrow margins. Reproducibility readiness: medium.
LingLan: multimodal diagnosis LLM for tongue and pulse diagnosis
Supervised fine-tuning with LoRA over multimodal data for TCM diagnosis; reports a 103.5% relative accuracy gain to 62.72%, which Evolve flags as promising but needing scrutiny given malformed metrics and inconsistent confusion-matrix counts. Reproducibility readiness: low.

Timeline

2026Co-authored TCMaster, introducing confidence-aware querying and workload-guided physical design for multi-source Traditional Chinese Medicine knowledge graphs (arXiv preprint).
2026Co-authored LingLan, a multimodal Traditional Chinese Medicine diagnosis LLM built with supervised fine-tuning and LoRA for tongue and pulse diagnosis (arXiv preprint).
2026Co-authored a structured-reasoning and knowledge-driven alignment pipeline for TCM prescription generation using chain-of-thought distillation and DPO (arXiv preprint).

Papers in the Evolve catalog

Frequent co-authors

Zheng Chen×3 Peiwu Qin×3 Zhicheng Du×2 Haoxuan Li×2 Yingshan Liang×1 Yuzhu Li×1 Zhongde Zhang×1 Lianshun Jin×1 Peiwu Qin×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric measures come from a limited OpenAlex-based snapshot; no citation counts, h-index, or venue-level records were provided.
No institution, country, or ORCID is available; Haoxuan Li is a common name, so this profile is name-resolved from public bibliographic data.
Output, impact, and activity dimensions lack sufficient underlying data and are left unscored rather than estimated.
All three papers are 2026 arXiv preprints reviewed by Evolve; no peer-reviewed venue records are represented in this catalog.