← back to paper
arxiv: 2604.10500 · 3 revisions
Visual Enhanced Depth Scaling for Multimodal Latent Reasoning