TY  - RPRT
TI  - Listen, Do Not Copy: Internalizing Audio-Grounded Scaffold Context for Robust Omni-Model Speech Understanding
AU  - Pengfei Zhang
AU  - Biao Tian
AU  - Tianxin Xie
AU  - Minghao Yang
AU  - Xiangang Li
AU  - Li Liu
PY  - 2026
UR  - https://arxiv.org/abs/2607.21943
ID  - 2607.21943
ER  -