Author(s): Dheeraj Sonkhla

Abstract: The use of Large Language Models (LLMs) has significantly transformed the digital healthcare. Pharmacy is a distinguished part of healthcare and recent researches in this domain have increased tremendously. This review article is focus on current advancements, techniques and data requirements of LLMs when dealing with pharmaceutical field. The paper analyzes the core techniques including Chain-of-Thought (CoT) prompting for structured reasoning, domain-specific fine-tuning, and Knowledge Graph integration to ensure interpretability. The Retrieval-Augmented Generation (RAG) are discussed in detail as it makes the LLM responses to mitigate hallucinations and improve accuracy. Despite these advancements, the review highlights critical limitations in current architectures. LLMs exhibit significant fragility to input noise and struggle with complex clinical guideline adherence. Notably, general-purpose models like GPT-4 demonstrated a 71% failure rate in detecting potential drug-drug interactions (pDDIs) compared to standard software, posing serious safety risks. The study concludes that while LLMs offer unprecedented opportunities for efficiency and information synthesis, they cannot yet function as autonomous agents. Safe implementation requires hybrid human-AI workflows, robust adversarial defenses, and harmonized regulatory frameworks to validate performance in high-stakes pharmaceutical environments.

DOI: 10.33545/26633582.2025.v7.i2d.237

Pages: 339-344 | Views: 81 | Downloads: 36

Download Full Article: Click Here