28/09/2026 10:53
|
Information Technology
Fine-tuning large language models using drunken text makes them more likely to leak secrets and answer harmful questions, new UNSW-led research shows.
UNSW researchers have shown that large...