{"status":"ok","resource":{"id":1781300,"copied_from_id":null,"created_at":"2025-01-27T15:00:31.088Z","external_ref":null,"favorite":null,"link_id":1682676,"meta":null,"notes":"TLDR;DeepSeek just made a breakthrough: you can train a model to match OpenAI o1-level reasoning using pure reinforcement learning (RL) without using labeled data (DeepSeek-R1-Zero). But RL alone isn’t perfect — it can lead to challenges like poor readability. A mix of methods in a multi-stage training fixes these (...","organization_id":null,"score":1655.3199466262038,"shared_at":"2025-01-27T15:00:31.087Z","title":"How DeepSeek-R1 Was Built, for Dummies","updated_at":"2025-01-27T15:06:04.095Z","user_id":33,"visibility":"shared","tefter_url":"https://tefter.io/bookmarks/1781300","tefter_edit_url":"https://tefter.io/bookmarks/1781300/edit","link":{"uri":"https://www.vellum.ai/blog/the-training-of-deepseek-r1-and-ways-to-use-it"}}}