Peiwu Qin

Research profile

Research summary

Peiwu Qin is a researcher at Guangdong Provincial Hospital of Chinese Medicine whose cataloged work sits at the intersection of Traditional Chinese Medicine (TCM) informatics, medical natural language processing, and knowledge-graph data management. The Evolve catalog holds three 2026 arXiv papers, on all of which Qin is the last-listed author (none flagged corresponding), consistent with a senior or supervisory role. LingLan builds a multimodal TCM diagnosis LLM with supervised fine-tuning and LoRA covering tongue and pulse diagnosis, reporting a 103.5% relative accuracy gain to 62.72%. Syndrome, Synergy, and Safety presents a structured-reasoning, knowledge-driven pipeline for TCM prescription generation using chain-of-thought distillation and DPO alignment, with medication safety and the Shi Ba Fan rules in scope. TCMMaster addresses confidence-aware querying and workload-guided physical design over a 221K-node multi-source TCM knowledge graph. Field distribution is an even three-way split (0.333 each), yielding a focus score of 0; collaboration scores 46 with 9 distinct coauthors; reproducibility averages 58 across 3 assessed papers, spanning low (LingLan), medium (prescription generation), and high (TCMaster). Output, impact, and activity are unscored for lack of citation and volume data.

Research keywords

Highlights

• TCMaster: confidence-aware querying and workload-guided design on a 221K-node TCM knowledge graph

  The 2026 systems paper pairs confidence-aware querying (Cypher rewriting, path ranking) with workload-guided physical design for multi-source TCM knowledge graphs. Evolve reads the gains as honest but modest, noting the 20-point question-answering advantage rests on a self-built benchmark; reproducibility is rated high.

• Structured reasoning and knowledge alignment for TCM prescription generation

  The 2026 pipeline combines chain-of-thought distillation and DPO alignment with knowledge-driven constraints covering medication safety and the Shi Ba Fan incompatibility rules. Evolve judges the pipeline's effects real, but the headline that a 7B model beats GPT-5 leans on inconsistent PQS reporting and tiny margins; reproducibility is medium.

• LingLan: multimodal LLM for TCM diagnosis with tongue and pulse data

  The 2026 diagnosis model applies supervised fine-tuning with LoRA over multimodal inputs including tongue and pulse diagnosis, reporting a 103.5% relative accuracy gain to 62.72%. Evolve flags malformed metrics and inconsistent confusion-matrix counts that need scrutiny; reproducibility is rated low.

Timeline

1. 2026Last-author arXiv paper on TCM prescription generation, combining structured reasoning, chain-of-thought distillation, and knowledge-driven DPO alignment.
2. 2026Last-author arXiv paper introducing LingLan, a multimodal TCM diagnosis LLM covering tongue and pulse diagnosis.
3. 2026Last-author arXiv systems paper TCMMaster on confidence-aware querying and workload-guided physical design for TCM knowledge graphs.

Papers in the Evolve catalog

• 2026-09-22

  Syndrome, Synergy, and Safety: Structured Reasoning and Knowledge-Driven Alignment for TCM Prescription Generation

• 2026-09-22

  LingLan: An Advancing Traditional Chinese Medicine Diagnosis LLM with Multimodal Data

• 2026-09-22

  TCMaster: Confidence-Aware Querying and Workload-Guided Physical Design for Multi-Source Traditional Chinese Medicine Knowledge Graphs

Frequent co-authors