Look and Think: Efficient Multimodal Reasoning via Modality-Decoupled Compression.

Xidi Cai, Junhao Zheng, Jingye Li, Boyuan Li, Shaowei Zhang, Qianli Ma

Browse the full ACL paper archive.