TY  - RPRT
TI  - Learning What Helps: Task-Aligned Context Selection for Vision Tasks
AU  - Jingyu Guo
AU  - Emir Konuk
AU  - Fredrik Strand
AU  - Christos Matsoukas
AU  - Kevin Smith
PY  - 2025
UR  - https://arxiv.org/abs/2512.00489
ID  - 2512.00489
ER  -