← back
arXivHongyu Ke, Jack Morris, Yongkang Liu, Satoshi Kitai, Kentaro Oguchi, Yi Ding, Haoxin WangWed, May 20, 2026, 8:36 AM PDT
score 13.3
3cites

Vision model learns dynamic image scanning instead of fixed patterns

Original: Deformba: Vision State Space Model with Adaptive State Fusion

Source: arxiv.org

Writing ELI5 summary…