CompoDistill: Attention Distillation for Compositional Reasoning in Multimodal LLMs

ArXi:2510.12184v2 Announce Type: replace Recently, efficient Multimodal Large Language Models (MLLMs) have gained significant attention as a solution to their high computational complexity, making them practical for real-world applications. In this regard, the knowledge distillation (KD) approach has emerged as a promising alternative, which transfers the rich visual and linguistic knowledge from a larger model (teacher) to a smaller model (student