一作ACL 2026Poster· Natural Language Processing
Forest Before Trees: Latent Superposition for Efficient Visual Reasoning
Yubo Wang · Juntian Zhang · Yichen WU · Yankai Lin · Nils Lukas · Yuhan Liu
摘要
While Chain-of-Thought empowers Large Vision-Language Models with multi-step reasoning, explicit textual rationales suffer from an information bandwidth bottleneck, where continuous visual details are discarded during discrete tokenization. Recent latent reasoning methods attempt to address this...