Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AyurParam: A State-of-the-Art Bilingual Language Model for Ayurveda

Mohd Nauman, Sravan Gvm, Vijay Devane, Shyam Pawar, Viraj Thakur, Kundeshwar Pundalik, Piyush Sawarkar, Rohit Saluja, Maunendra Desarkar, Ganesh Ramakrishnan

TL;DR

AyurParam addresses the gap where general LLMs struggle with Ayurveda by training AyurParam-2.9B, a bilingual model on a meticulously curated corpus of classical texts and clinical guidance. Using domain-adaptive instruction tuning and knowledge-grounded QA, it achieves state-of-the-art performance among 1.5–3B models on BhashaBench-Ayur and competitive results with larger models, underscoring the value of authentic domain supervision. The work presents a comprehensive data-generation and evaluation pipeline, highlights Hindi-English performance disparities, and outlines practical steps toward clinically validated, safe, and scalable Ayurvedic AI. It also emphasizes ethical data use and the need for diverse, contemporary sources to improve generalization and real-world utility.

Abstract

Current large language models excel at broad, general-purpose tasks, but consistently underperform when exposed to highly specialized domains that require deep cultural, linguistic, and subject-matter expertise. In particular, traditional medical systems such as Ayurveda embody centuries of nuanced textual and clinical knowledge that mainstream LLMs fail to accurately interpret or apply. We introduce AyurParam-2.9B, a domain-specialized, bilingual language model fine-tuned from Param-1-2.9B using an extensive, expertly curated Ayurveda dataset spanning classical texts and clinical guidance. AyurParam's dataset incorporates context-aware, reasoning, and objective-style Q&A in both English and Hindi, with rigorous annotation protocols for factual precision and instructional clarity. Benchmarked on BhashaBench-Ayur, AyurParam not only surpasses all open-source instruction-tuned models in its size class (1.5--3B parameters), but also demonstrates competitive or superior performance compared to much larger models. The results from AyurParam highlight the necessity for authentic domain adaptation and high-quality supervision in delivering reliable, culturally congruent AI for specialized medical knowledge.

AyurParam: A State-of-the-Art Bilingual Language Model for Ayurveda

TL;DR

AyurParam addresses the gap where general LLMs struggle with Ayurveda by training AyurParam-2.9B, a bilingual model on a meticulously curated corpus of classical texts and clinical guidance. Using domain-adaptive instruction tuning and knowledge-grounded QA, it achieves state-of-the-art performance among 1.5–3B models on BhashaBench-Ayur and competitive results with larger models, underscoring the value of authentic domain supervision. The work presents a comprehensive data-generation and evaluation pipeline, highlights Hindi-English performance disparities, and outlines practical steps toward clinically validated, safe, and scalable Ayurvedic AI. It also emphasizes ethical data use and the need for diverse, contemporary sources to improve generalization and real-world utility.

Abstract

Current large language models excel at broad, general-purpose tasks, but consistently underperform when exposed to highly specialized domains that require deep cultural, linguistic, and subject-matter expertise. In particular, traditional medical systems such as Ayurveda embody centuries of nuanced textual and clinical knowledge that mainstream LLMs fail to accurately interpret or apply. We introduce AyurParam-2.9B, a domain-specialized, bilingual language model fine-tuned from Param-1-2.9B using an extensive, expertly curated Ayurveda dataset spanning classical texts and clinical guidance. AyurParam's dataset incorporates context-aware, reasoning, and objective-style Q&A in both English and Hindi, with rigorous annotation protocols for factual precision and instructional clarity. Benchmarked on BhashaBench-Ayur, AyurParam not only surpasses all open-source instruction-tuned models in its size class (1.5--3B parameters), but also demonstrates competitive or superior performance compared to much larger models. The results from AyurParam highlight the necessity for authentic domain adaptation and high-quality supervision in delivering reliable, culturally congruent AI for specialized medical knowledge.

Paper Structure

This paper contains 28 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data Preparation
  4. Taxonomy of Ayurvedic Domains
  5. Corpus Collection and License Governance
  6. OCR and Post-OCR Processing
  7. Knowledge-Grounded Data Generation
  8. Ethical and Legal Considerations
  9. Training
  10. Training Data
  11. Conversational Alignment Data.
  12. Training Setup
  13. Model Performance
  14. Conclusion
  15. Limitations
  16. ...and 13 more sections

Figures (3)

  • Figure 1: Data preparation pipeline
  • Figure 2: Language-wise distribution of the collected corpus across Sanskrit, Hindi, Marathi, and English sources
  • Figure 3: QA-Distribution