§ feed · storyline

Reasoning models struggle to control their chains of thought, and that’s good

OpenAI introduces CoT-Control, finding that reasoning models cannot reliably steer their own chains of thought, which the lab says supports monitorability as an AI safety measure.

Mar 5 · 11:00:00 · primary fetch1 sourceupdated Mar 5 · 11:00:00

OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety safeguard.

read full article on openai.com ↗

§ sources1 publication · timeline below

openai.comReasoning models struggle to control their chains of thought, and that’s goodprimary11:00:00