OpenAI researcher says Fable 5.1 far less monitorable than Astra
OpenAI researcher tomekkorbak said Fable 5.1 and Mythos 5.1 are significantly less monitorable than Astra, claiming their CoT monitor could have caught a Hugging Face-like incident where Anthropic's could not, and noting evaluation setups often differ from production monitoring.
2026-09-16 ~ 2026-09-16 · 4 related posts
- OpenAI researcher: Fable 5.1 and Mythos 5.1 are far less monitorable than Astra — tomekkorbak · 2026-09-16
- CoT monitorability evals use CoT-only monitors, but prod monitors see actions too — tomekkorbak · 2026-09-16
- Researcher: Our CoT Monitors Would've Flagged the Hugging Face Incident — tomekkorbak · 2026-09-16
1 near-duplicate retellings: tomekkorbak