New Talk: Building Olmo 3 Think

<p>It’s finally here! The public (and most complete) version of my talk covering every stage of the process to build <a href="https://www.interconnects.ai/p/olmo-3-americas-truly-open-reasoning" target="_blank">Olmo 3</a> Think (<a href="https://docs.google.com/presentation/d/1rhXAPyaflqi5cBS6d6IDiNIvLIBxiOpVb1sfK3hNMmo/edit?usp=sharing" target="_blank">slides</a> are available). I’ve been giving this, improving it, and getting great feedback at other venues such as The Conference on Language Modeling (COLM) & The PyTorch Conference.This involves changes and new considerations of every angle of the stack, from pretraining, evaluation, and of course post-training.</p><p>Most of the talk focuses on reinforcement learning infrastructure and evaluating reasoning models, with quick comments on every training stage. I hope you enjoy it, and let us know what to improve in the future!</p><p><strong>Chapters</strong></p><p>* 00:00:00 Introduction</p><p>* 00:06:30 Pretraining Architecture</p><p>* 00:09:25 Midtraining Data</p><p>* 00:11:08 Long-context Necessity</p><p>* 00:13:04 Building SFT Data</p><p>* 00:20:05 Reasoning DPO Surprises</p><p>* 00:24:47 Scaling RL</p><p>* 00:41:05 Evaluation Overview</p><p>* 00:48:50 Evaluation Reflections</p><p>* 01:00:25 Conclusions</p><p>Here’s the YouTube link:</p><p></p> <br /><br />This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit <a href="https://www.interconnects.ai/subscribe?utm_medium=podcast&#38;utm_campaign=CTA_2">www.interconnects.ai/subscribe</a>