MAC-RRG: Iterative Multi-Agent Collaboration for X-ray Radiology Report Generation

MAC-RRG implements a closed-loop optimization paradigm for X-ray radiology report generation using iterative multi-agent collaboration. The framework decouples tasks by using a vision encoder and LLM for initial reporting, followed by dynamic knowledge updating through multi-agent reasoning to mitigate medical hallucinations common in static knowledge graph methods.