TY  - RPRT
TI  - Not All Tokens See Equally: Perception-Grounded Policy Optimization for Large Vision-Language Models
AU  - Zekai Ye
AU  - Qiming Li
AU  - Xiaocheng Feng
AU  - Ruihan Chen
AU  - Ziming Li
AU  - Haoyu Ren
AU  - Kun Chen
AU  - Dandan Tu
AU  - Bing Qin
PY  - 2026
UR  - https://arxiv.org/abs/2604.01840
ID  - 2604.01840
ER  -