Poster Thu, Sep 10, 2026 • 7:30 AM – 9:30 AM PDT ExHall # 7

Semantic-Geometric Dual Compression: Training-Free Visual Token Reduction for Ultra-High-Resolution Remote Sensing Understanding

yueying li ⋅ Fengxiang Wang ⋅ Yan Li ⋅ Mingshuo Chen ⋅ Mengying Zhao ⋅ Long Lan

Keywords: Video Understanding, Generation and Multimodal Learning