AI papers News and Analysis

2 verified stories covering AI papers, product updates and industry developments.

AI Safety 2026-10-11

Redwood uses distillation to audit models; confession hits 84%

Redwood Research's new paper distills untrusted models into same-base students to make them confess hidden traits, lifting one trait's confession rate from 22% to 84%.

#AI alignment #AI papers

arXiv 2026-05-18

arXiv threatens one-year bans for unchecked AI slop

The preprint server says papers with obvious unreviewed LLM output, hallucinated citations or leftover model comments can trigger a year-long submission ban.

#AI papers #Academia #AI hallucinations