Compose and Fuse: Revisiting the Foundational Bottlenecks in Multimodal Reasoning

arXiv:2509.23744v3 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) promise enhanced reasoning by integrating diverse inputs such as text, vision, and audio. Yet cross-modal reasoning remains underexplored, with conflicting reports on whether added modalities help or...