Research2h ago
Apple Studies Multimodal LLM Alignment
A new study from Apple examines how preference alignment reduces hallucinations in multimodal models. These systems often produce responses that contradict visual input rather than just stating false facts. The research focuses on forcing models to prioritize image information over internal biases. This provides a technical framework for improving visual grounding in MLLMs.