fp.
newest
Open in hackernews
Even GPT-5.2 Can't Count to Five: Zero-Error Horizons in Trustworthy LLMs
https://arxiv.org/abs/2601.15714
4
•
daigoba66
•
1h ago