The Olmo 3 pipeline and training reasoning models
A post published on December 15, 2025 describes SFT with curated traces, followed by DPO and RLVR using group-relative policy optimization.
Source: Pipeline de treinamento do Olmo 3 com SFT, DPO e RLVR (x.com). Text prepared with AI from this source.
What happened and what to do
Published on December 15, 2025, the post describes the Olmo 3 training pipeline: SFT on curated reasoning traces, followed by preference tuning with DPO and RLVR using group-relative policy optimization. According to the post, combining SFT and DPO is a better starting point for RL than using SFT alone or applying RL directly to the base model. The text highlights how data curation and preference tuning influence the training of reasoning models.
A company training or adapting models could implement a controlled workflow to organize curated data, record tuning stages, and evaluate outcomes across different training strategies. The consultancy can help define metrics, automate experiment collection and monitoring, and structure execution on the company’s own infrastructure.
How the consultancy can help
Wendelmaques can assess your data and training goals, define an implementation scope for curation, tuning, and evaluation pipelines, and support the solution’s operation and monitoring.
Next step
Send a short description of your model training or adaptation case to receive a scoped proposal.
Let's assess your project
I reply with scope and a proposal.