Med-V1: Small Language Models for Zero-shot and Scalable Biomedical Evidence Attribution
Researchers have developed Med-V1, a family of small language models designed for efficient biomedical evidence attribution. These three-billion-parameter models, trained on synthetic data, significantly outperform their base models and rival frontier LLMs like GPT-5 on biomedical benchmarks. Med-V1 was used to study hallucinations in LLM-generated answers and identify misattributions in clinical guidelines, highlighting potential public health impacts. AI
IMPACT Provides an efficient, lightweight alternative to frontier LLMs for critical biomedical evidence attribution and verification tasks.