Beta
Train Profile
Train optimal weights for a function by providing example inputs with expected outputs. ObjectiveAI learns the best ensemble weights to minimize loss on your training data.
Function
Loading functions...
Training Parameters
A single value in [0, 1] — the desired function output.
Dataset
1 itemItem 1
Training Results
When training completes, this section will show:
- Learned Weights — Optimized weights for each LLM in the ensemble
- Fitting Stats — Loss, rounds, executions, and errors
- Profile Export — Copy as profile JSON for GitHub hosting
How Profile Training Works
ObjectiveAI doesn't fine-tune LLMs—it learns optimal weights over fixed models. This is faster, more interpretable, and doesn't require model access.
Training Process:
- You provide a dataset of inputs with target outputs
- ObjectiveAI executes the function n times per dataset item
- A loss function measures how far outputs are from targets
- Weights are adjusted to minimize loss across multiple starts and rounds
- The final weights are saved as a Profile
The resulting Profile can be hosted on GitHub and used with Functions.Executions.create() to apply the learned weights.