Researchers have evaluated the use of large language models (LLMs) to generate behavioral patterns for epidemiological modeling, specifically focusing on hepatitis C virus-related behaviors among people who inject drugs. The study found that LLMs could recover directional associations from literature with 46-68% accuracy, and when directly prompted, this accuracy increased to 82-93%. However, the alignment between LLM-generated behavioral relationships and epidemiological evidence was only partial, with mean absolute differences from agent-based models and survey correlations ranging from 0.11-0.18. The findings suggest that while LLMs show promise, more systematic evaluation is required before their outputs can be reliably incorporated into transmission dynamics simulations. AI
IMPACT LLMs demonstrate potential for generating behavioral data in epidemiological studies, though further validation is needed for direct use in transmission simulations.
RANK_REASON The cluster contains a research paper published on arXiv detailing an evaluation framework for LLM-generated patterns in epidemiological modeling. [lever_c_demoted from research: ic=1 ai=1.0]
- agent-based model
- arXiv
- epidemiological modeling
- Hepatitis C virus
- Hugging Face
- large language model
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
