ViCA: Efficient Multimodal LLMs with Vision-Only Cross-Attention

Documenti analoghi

• StreamingThinker: Large Language Models Can Think While Reading
  di: Tong, Junlong, et al.
  Pubblicazione: (2025)
• $\mathcal{V}isi\mathcal{P}runer$: Decoding Discontinuous Cross-Modal Dynamics for Efficient Multimodal LLMs
  di: Fan, Yingqi, et al.
  Pubblicazione: (2025)
• What Do Visual Tokens Really Encode? Uncovering Sparsity and Redundancy in Multimodal Large Language Models
  di: Fan, Yingqi, et al.
  Pubblicazione: (2026)
• Decoupling KL and Trajectories: A Unified Perspective for SFT, DAgger, Offline RL, and OPD in LLM Distillation
  di: Zhao, Anhao, et al.
  Pubblicazione: (2026)
• HiDrop: Hierarchical Vision Token Reduction in MLLMs via Late Injection, Concave Pyramid Pruning, and Early Exit
  di: Wu, Hao, et al.
  Pubblicazione: (2026)