Paper and code released: Reward Misspecification Degrades CoT Monitorability
johnhewtt · x · 2026-10-20
Resource thread for the study above: the paper "Reward Misspecification Degrades Chain-of-Thought Monitorability" is out, with an Apache-2.0 GitHub repo providing datasets, configs and reproduction scripts (built on tinker-cookbook==0.5.7, with Tinker, OpenAI and W&B integration). The lab is also seeking funding partners for rigorous safety research.
Related event: Study: Reward Misspecification Makes Chain-of-Thought Unmonitorable(9 posts)→
More from Safety
- ByteDance, Alibaba, Tencent order 3M+ Nvidia RTX PRO 5500 cards worth $35B, Reuters reports — SumitGup · 2026-10-21
- CVSS 10.0 RCE vulnerability disclosed in PHP core, patches only for PHP 8 — cyb3rops · 2026-10-21
- Antigen launches Assets: a security agent that maps your infrastructure and triages new CVEs — alishbaimran_ · 2026-10-21
- GitHub HQ Hosts Open Source AI + Policy Talk with Mozilla Co-founder Mitchell Baker — kevinsxu · 2026-10-21
- Researcher Finds Two Windows 0-Days, First in 76 Minutes with a Frontier Model — moyix · 2026-10-21
- Stanford's Noah Goodman shares interactive AI safety concept map — stanfordnlp · 2026-10-21