Ryze: Evidence-Enriched Data Synthesis from Biomedical Papers
Researchers have developed Ryze, an automated system designed to create a specialized vision-language model (VLM) for biomedical research by synthesizing evidence-enriched training data from scientific papers. This system extracts and structures information from figures, tables, and text, overcoming limitations of previous methods that relied on costly expert annotation or lost evidence context. The resulting BioVLM-8B model, trained using Ryze for under $200, achieved a 48.0% weighted accuracy on the LAB-Bench benchmark, surpassing both its base model and GPT-5.2. AI
IMPACT Enables more accurate biomedical research by improving VLM capabilities with structured, evidence-rich data.