Evidence item

An OpenAI researcher states RLSlow was named after Thinking, Fast and Slow: RL would teach models 'slow' thinking. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 7200

Source: Trapit Bansal Twitter profile · locator {} · excerpt hash sha256:23d17ab5da8e284df5d55d5ccbdaefa434c36ce27a03f7a87f6b89e82b86b0a9

Fun fact: RLSlow was named after Thinking, Fast and Slow. The idea was that language models already had a kind of 'fast' thinking, producing an answer immediately, and that we could use RL to teach them 'slow' thinking: deliberate, multi-token reasoning that spends more compute working through a problem.