RSS

Pablo Bernabeu-Pérez

Karma: 70

Ensem­ble mon­i­tor­ing for AI con­trol: di­verse sig­nals out­weigh more compute

Yejun Y., Sam Tetef, eugenekoran, Benjamin Arnav and Pablo Bernabeu-Pérez
31 May 2026 1:21 UTC
12 points
0 comments7 min readLW link

Un­faith­ful Rea­son­ing Can Fool Chain-of-Thought Monitoring

Benjamin Arnav, Pablo Bernabeu-Pérez, Tim Kostolansky, HanneWhitt, Nathan Helm-Burger and Mary Phuong
2 Jun 2025 19:08 UTC
78 points
17 comments3 min readLW link