AgriConflict

New CAER framework improves multimodal LLM reliability by routing conflicting evidence

Researchers have developed CAER, a novel framework designed to address conflicts between textual claims and visual evidence in Multimodal Large Language Models (MLLMs). CAER employs a span-grounded evidence router to id…