A new paper introduces MeSH-Rel-4K, a dataset of 4,000 semantic relationships from the Medical Subject Headings (MeSH) to evaluate resource-efficient Large Language Models (LLMs) in biomedical ontology generation. The study benchmarks five open-source LLMs with up to 9 billion parameters, testing standard prompting, Chain-of-Thought, and fine-tuning adaptation strategies. Results show that fine-tuning significantly improves performance, increasing the average F1-score by 34.1 percentage points, demonstrating an effective automated method for creating specialized biomedical ontologies. AI
IMPACT Fine-tuning smaller LLMs effectively for specialized biomedical ontologies could accelerate knowledge organization in the field.
RANK_REASON The cluster describes a research paper presenting a new dataset and benchmarking LLMs for a specific task. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
