Experiment brief

Experiment name

Opening your draft…

Only on this browser and site. Export a copy to move it; this address isn’t shared with your account.

What should your agent do better?

Describe a specific task, behavior, or recurring mistake.

Your next stepTrain a modelEvaluate a model

Next, choose existing tasks from a published dataset. This brief does not generate tasks or define their scoring. You’ll review the run inputs before confirming launch.

Add model, examples and success criteria (optional)

Model or checkpoint (optional)

Examples (optional)

Brief notes, not a published dataset.

How will you know it improved?

Planning notes only. The selected dataset defines the scoring used by a run.

New experiment

Recent drafts in this browser

No other saved drafts in this browser.

Learn about tune.new