TY  - RPRT
TI  - Confident, Calibrated, or Complicit: Safety Alignment and Ideological Bias in LLM Hate Speech Detection
AU  - Sanjeeevan Selvaganapathy
AU  - Mehwish Nasim
PY  - 2026
UR  - https://arxiv.org/abs/2509.00673
ID  - 2509.00673
ER  -