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 item
Item 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:

  1. You provide a dataset of inputs with target outputs
  2. ObjectiveAI executes the function n times per dataset item
  3. A loss function measures how far outputs are from targets
  4. Weights are adjusted to minimize loss across multiple starts and rounds
  5. 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.