Causal distillation teaches small AI models to use visual evidence

Original: MCD: Causal Distillation of Multimodal In-Context Learning in Large Vision-Language Models

Source: arxiv.org ↗