TY  - RPRT
TI  - Even GPT-5.2 Can't Count to Five: The Case for Zero-Error Horizons in Trustworthy LLMs
AU  - Ryoma Sato
PY  - 2026
UR  - https://arxiv.org/abs/2601.15714
ID  - 2601.15714
ER  -