DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs

Nov-15-2025, 05:21:37 GMT–Neural Information Processing Systems

These gains are particularly pronounced on high-resolution tasks, e.g ., 4.2, 11.0, and 4.0 improvements on TextVQA, DocVQA, and InfoVQA compared to LLaV A-1.5-7B, respectively.

large language model, machine learning, natural language, (19 more...)

Neural Information Processing Systems

Nov-15-2025, 05:21:37 GMT

Conferences PDF

Country:
- Asia > China
  - Shanghai > Shanghai (0.04)
- North America > United States
  - California > Santa Clara County > Palo Alto (0.04)

Genre:
- Research Report > Experimental Study (0.93)

Technology:
- Information Technology
  - Artificial Intelligence
    - Machine Learning > Neural Networks
      - Deep Learning (1.00)
    - Natural Language
      - Chatbot (0.93)
      - Large Language Model (1.00)
    - Vision (1.00)
  - Sensing and Signal Processing > Image Processing (0.94)