Doubt on training

Question

Doubt on training

Closed this issue 7 months ago · 3 comments

Is a training on LLM done for OptiMUS or was gpt-4 equipped to give such outputs?
Is the dataset created used to evaluate the performance of LLM or used to train the LLM?
Can you give some insight into this?

Answer 1 · 2024-02-15T18:21:04.000Z

Hi
It uses gpt-4 and no training/fine-tuning is done for this version. The dataset was used for evaluation, but could be used for training too.

Answer 2 · 2024-02-16T04:36:02.000Z

So for all the steps that happen in the UI (OptiMUS website) is all by prompts to the gpt-4?

Answer 3 · 2024-02-17T00:14:32.000Z

in the version that is currently up yes. We're working on tuning, and will share the results in our next paper.