LLM4Sweat: A Trustworthy Large Language Model for Hyperhidrosis Support
Wenjie Lin, Jin Wei-Kocsis · 2025
Read LLM4Sweat: A Trustworthy Large Language Model for Hyperhidrosis Support by Wenjie Lin, Jin Wei-Kocsis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-15192.txt.
Introduction
Wenjie Lin and Jin Wei-Kocsis build a language model for excess sweat support in a 2025 paper. The disorder, hyperhidrosis, makes sweat far past need and hits two to three percent of people. It strains body ease and social life. Rare ills lack the clean data sets that fine tune needs. No past work tuned a model for this ill. Their system, LLM4Sweat, runs in three stages to close that gap.
First, a top model drafts fake case tales from curbed open data to build a fair question set. Next, an open base model tunes on that set for finds, care picks, and kind support. Last, clinic and mind experts grade the replies, and sound ones feed back to the set. Tests beat base lines. The authors free the full frame as a path for rare ill support.
Catalog record
| CREATOR | Wenjie Lin, Jin Wei-Kocsis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-15192
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
LLM4Sweat: A Trustworthy Large Language Model for Hyperhidrosis Support Wenjie Lin, Jin Wei-Kocsis Abstract While large language models (LLMs) have shown promise in healthcare, their application for rare medical conditions is still hindered by scarce and unreliable datasets for fine-tuning. Hyperhidrosis, a disorder causing excessive sweating beyond physiological needs, is one such rare disorder, affecting 2-3% of the population and significantly impacting both physical comfort and psychosocial well-being. To date, no work has tailored LLMs to advance the diagnosis or care of hyperhidrosis. To address this gap, we present LLM4Sweat, an open-source and domain-specific LLM framework for trustworthy and empathetic hyperhidrosis support. The system follows a three-stage pipeline. In the data augmentation stage, a frontier LLM generates medically plausible synthetic vignettes from curated open-source data to create a diverse and balanced question-answer dataset. In the fine-tuning stage, an open-source foundation model is fine-tuned on the dataset to provide diagnosis, personalized treatment recommendations, and empathetic psychological support. In the inference and expert evaluation stage, clinical and psychological specialists assess accuracy, appropriateness, and empathy, with validated responses iteratively enriching the dataset. Experiments show that LLM4Sweat outperforms baselines and delivers the first open-source LLM framework for hyperhidrosis, offering a generalizable approach for other rare diseases with similar data and trustworthiness challenges. Introduction Figure 1: Overview of our proposed LLM4Sweat framework. Hyperhidrosis, characterized by excessive sweating beyond thermoregulatory needs, affects approximately 2-3% of the population and can significantly impair both physical comfort and psychosocial well-being ( Strutton et al. 2004 ; Solish, Benohanian, and Kowalski 2007 ) . Current therapeutic protocols span from potent topical antiperspirants to advanced procedural interventions such as botulinum toxin injections, microwave thermolysis (miraDry), iontophoresis, and endoscopic thoracic sympathectomy ( Hong, Lupin, and O’Shaughnessy 2012 ; Gregoriou et al. 2019 ; Martínez-Hernández et al. 2024 ) . Despite the availability of these treatments, reliable and personalized medical support for diagnosis, treatment selection, and psychosocial management remains limited and unregulated ( Parashar, Adlam, and Potts 2023 ) . Notably, the majority of cases represent primary hyperhidrosis, which is inherited and characterized by focal symptoms and has been linked to several genetic loci ( Henning, Pedersen, and Jemec 2019 ) . Beyond clinical progress, parallel efforts in engineering have sought to improve diagnosis and daily management. Examples include polymer‐based insole designs using additive manufacturing for plantar hyperhidrosis management ( Camargo et al. 2024 ) , rectangular patch antenna electrolyte sensors for sweat composition analysis ( Sakhawat, Islam, and Reza 2024 ) , and a wearable system for continuous sweat level monitoring ( Bellapukonda, Mohan, and Sahu 2023 ) . Also, daily management devices are proposed, alongside systematic engineering reviews of diagnostic and intervention technologies ( Lin and Fang 2022 ) , illustrating the potential for integrating engineering innovations into hyperhidrosis care. However, they remain largely decoupled from intelligent reasoning systems that can deliver holistic diagnosis, treatment guidance, and psychosocial support. LLMs have shown promising capabilities in a range of healthcare applications, including diagnosis support ( Singhal et al. 2023 ) , treatment planning ( Zhao et al. 2025 ) , and patient counseling ( Qiu and Lan 2024 ) . However, their application adapted to hyperhidrosis and similar rare disorders remains largely absent. Unlike common conditions with abundant datasets, hyperhidrosis suffers from scarce high-quality data and fragmented, unreliable online information. These gaps hinder the trustworthy use of LLMs in clinical care and underscore the need for data-centric frameworks that can turn limited resources into reliable medical knowledge. To address these limitations, we present LLM4Sweat, the first open-source LLM framework designed for trustworthy hyperhidrosis support. It overcomes the aforementioned challenges through a three-stage pipeline. Our work makes three key contributions: (1) we construct the hyperhidrosis question-answer dataset from open-source data and generate medically plausible synthetic vignettes by a frontier LLM; (2) we introduce a closed-loop fine-tuning and expert-in-the-loop evaluation pipeline that adapts open-source lightweight LLM to perform three integrated tasks, provide diagnosis, recommend treatment, and psychological support, within a unified framework; and (3) experiments show that our framework outperforms baselines and demonstrates strong potential for clinical deployment. Beyond addressing a critical gap in LLM adaptation for hyperhidrosis, it also offers a transferable solution to tackle data scarcity and trustworthiness challenges across other rare medical conditions. Related Work Clinical Diagnosis and Treatment of Hyperhidrosis The diagnosis of hyperhidrosis is primarily clinical, relying on patient history and physical examination, though these assessments are often subjective in practice ( Henning et al. 2021 ) . Once identified, treatment typically follows a stepwise approach, beginning with topical aluminum chloride hexahydrate and advancing to oral anticholinergics, iontophoresis, botulinum toxin injections, or device-based options such as miraDry ( Hong, Lupin, and O’Shaughnessy 2012 ; Gregoriou et al. 2019 ) . For severe and refractory cases, surgical intervention with endoscopic thoracic sympathectomy may be considered, although it carries a significant risk such as compensatory sweating ( Martínez-Hernández et al. 2024 ) . Engineering Solutions for Hyperhidrosis Recent work has highlighted engineering-driven diagnostic and monitoring innovations. Sakhawat et al. designed an antenna-based electrolyte sensor using Whatman filter paper to detect sodium and chloride concentrations in sweat for palmoplantar hyperhidrosis patients ( Sakhawat, Islam, and Reza 2024 ) . 3D-printed thermoplastic polyurethane insoles have been developed to manage plantar hyperhidrosis by improving comfort and reducing moisture retention ( Camargo et al. 2024 ) . Hyun et al. leveraged machine learning models to classify hyperhidrosis types and predict compensatory hyperhidrosis level after surgery with high accuracy ( Hyun et al. 2023 ) . Bellapukonda et al. applied logistic regression to predict sweat levels with high accuracy based on environmental, physiological, and demographic variables, ( Bellapukonda, Mohan, and Sahu 2023 ) . Lin and Fang ( Lin and Fang 2022 ) provided a comprehensive review of current and potential hyperhidrosis interventions from both medical and engineering perspectives, proposing smart wearable devices for daily management. Synthetic Data in Healthcare Synthetic data generation has emerged as a solution for medical domains with limited annotated datasets ( Peng et al. 2023 ) . GAN-based models have also been used to expand rare disease datasets ( Kazeminia et al. 2020 ) . Diffusion models have been successfully applied to dermatological imaging tasks, producing high-fidelity synthetic skin lesion data to improve classifier robustness ( Farooq et al. 2024 ) . For text-based healthcare tasks, LLM-driven augmentation has improved performance in low-resource clinical NLP applications ( Barr et al. 2025 ) . LLMs in Medical Applications Recent work has shown that state-of-the-art LLMs such as GPT and Llama achieve competitive performance in clinical reasoning, treatment recommendation, and patient communication ( Singhal et al. 2023 ) . LLMs have also been used for empathetic patient support ( Alanezi 2024 ) . Psychological distress is also a key challenge in rare conditions such as hyperhidrosis ( Solish, Benohanian, and Kowalski 2007 ) , underscoring the need for both accurate medical guidance and empathetic support. Building on these advances, our framework leverages a frontier LLM to generate medically plausible vignettes from open-source hyperhidrosis-related data, converting scarce and noisy real-world resources into a structured and balanced dataset. This enriched dataset enables reliable fine-tuning, equipping the model to provide trustworthy diagnostic reasoning, personalized treatment recommendations, and empathetic counseling. Furthermore, in the inference and expert evaluation