{"id":"tp-48181","type":"article-newspaper","title":"Blocking Common Opening Tokens and Adding an Example Raised Reasoning Compliance in Two AI Models.","author":"Simon Yoon","publisher":"TOKENPOST Inc.","url":"https://www.tokenpost.com/news/technology/48181","datePublished":"2026-11-04T03:19:09.000Z","dateModified":"2026-11-04T03:19:09.000Z","language":"en","abstract":"Blocking “Okay” and “Alright” while providing an example raised average written-reasoning compliance by 41.2 percentage points for Qwen3-32B and 34.5 percentage points for OLMo-3.1-32B-Think; the scores did not measur…","section":"Technology","references":[{"title":"research on monitoring models’ intermediate reasoning","url":"https://www.tokenpost.com/news/technology/28442"}]}