
Train and Promote a Model
Training changes model weights. It should be the result of a measured decision, not an automatic response to a failed chat. OpenPond keeps training separate from Refiner and Harness updates so the higher-cost, less-reversible change has its own data, evaluation, budget, and activation gates.

Prepare the evidence first
Start from reviewed chats or another approved source, then create a Taskset. Check that every example has understandable lineage, the graders fit the job, and a frozen evaluation split is not reused as training data.
Choose the workflow and adapter after that work is ready:
| Workflow | Good fit |
|---|---|
| SFT | Learn desired responses from reviewed demonstrations. |
| RL | Learn from rewards on fresh responses. |
| Decision model | Train or score allowed choices from structured state. |
LoRA learns low-rank updates, Engram learns conditional memory, and Neuron training learns selected weight columns. See the adapter guide for setup.
Review the exact run
In the web app, open Models, choose New model, then its workflow, adapter and starting model. The Desktop images below illustrate the earlier setup; they are not the current web chooser. Confirm the Taskset revision and hash, starting model, method, compute destination, retention, quote, and maximum spend before selecting Start training.

A successful job means the run completed. It does not prove the candidate should be used.
Evaluate and promote
Compare the candidate against the frozen baseline. Inspect failed graders and qualitative behavior. If the evidence does not meet the promotion threshold, keep the candidate unbound, revise the Taskset, and run another candidate.

An explicit promotion gate protects the active model binding and makes rollback clear. Follow the full Desktop training guide for the end-to-end workflow.