TY  - RPRT
TI  - Proxy reliance in large language model decisions is uncalibrated to predictive evidence
AU  - Zengqing Wu
AU  - Chuan Xiao
PY  - 2026
UR  - https://arxiv.org/abs/2608.22887
ID  - 2608.22887
ER  -