Early work on monitorability evaluations - METR
We show preliminary results on a prototype evaluation that tests monitors' ability to catch AI agents doing side tasks, and AI agents' ability to bypass this monitoring.
Early work on monitorability evaluations - METR Our Work Research Notes Updates Risk Assessment About Donate Careers Search --> Our Work Research Notes Updates Risk Assessment About Donate Careers Menu × Early work on monitorability evaluations CONTRIBUTORS Megan Kinniment , Seraphina Nix , Thomas Broadley , Hjalmar Wijk , and Neev Parikh DATE January 22, 2026 SHARE Copy Link Citation BibTeX Citation × @misc { metr-2026-early-work-on-monitorability-evaluations , title = {Early work on monitorability evaluations} , author = {Megan Kinniment, Seraphina Nix, Thomas Broadley, Hjalmar W
saved by
related reading
- MALT: A Dataset of Natural and Prompted Behaviors That Threaten Eval Integrity - METRmetr.org
- Frontier Risk Report (February to March 2026) - METRmetr.org
- [2512.18311] Monitoring Monitorabilityarxiv.org
- Demystifying evals for AI agents \ Anthropicanthropic.com
- Measuring AI Ability to Complete Long Tasks - METRmetr.org
- metr.org/risk-report-feb-mar-2026.pdf#page=33.47metr.org
- [2507.11473] Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safetyarxiv.org
- Summary of METR's predeployment evaluation of GPT-5.6 Solmetr.org
- Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscationarxiv.org
- Metagaming matters for training, evaluation, and oversightalignment.openai.com
- Oversight Assistants: Turning Compute into Understandingbounded-regret.ghost.io
- How independent researchers could investigate AI propensities after misalignment incidents - METRmetr.org