TY  - RPRT
TI  - Local Evidence, Global Failure: Diagnosing Speech-Vision Temporal Composition in Audio-Visual Language Models
AU  - Jaehong Myung
AU  - Seunghun Oh
AU  - Unsang Park
PY  - 2026
UR  - https://arxiv.org/abs/2610.25352
ID  - 2610.25352
ER  -