[공지사항]을 빙자한 안부와 근황 
Show more

Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

RAG-based Architectures for Drug Side Effect Retrieval in LLMs

Created by
  • Haebom

Author

Shad Nygren, Pinar Avci, Andre Daniels, Reza Rassol, Afshin Beheshti, Diego Galeano

Outline

In this paper, we propose a novel approach to utilize LLM for adverse drug reaction detection and analysis, despite the __T2606__ of large-scale language models (LLMs). To address the problems of traditional LLMs, such as black-box learning data dependency, hallucination phenomenon, and lack of domain-specific knowledge, we propose two architectures, RAG and GraphRAG, that integrate comprehensive adverse drug reaction knowledge into the Llama 3 8B language model. Experimental results using a dataset of 19,520 adverse drug reaction associations show that GraphRAG achieves near-perfect accuracy in adverse drug reaction detection. This provides an accurate and scalable solution, which represents a significant advancement in the application of LLMs in the field of pharmacovigilance.

Takeaways, Limitations

•
Takeaways:
◦
Suggesting the possibility of improving the accuracy of drug side effect detection and analysis using LLM
◦
Confirmation of the possibility of building an efficient and accurate drug side effect information retrieval system through GraphRAG architecture
◦
Presenting new possibilities for the use of LLM in the field of drug surveillance
◦
Contribute to building an accurate and scalable drug side effect information system
•
Limitations:
◦
Further verification of the generalization performance of the proposed model is needed.
◦
Additional research is needed on applicability and stability in real clinical settings.
◦
Consideration should be given to the limitations and biases of the dataset used.
◦
It is not a complete solution to the limitations of the Llama 3 8B model itself (e.g. potential for hallucinations).
👍
Made with Slashpage