When you call the OpenAI API, you can ask for log-probabilities alongside the generated text. These logprobs tell you how confident the model was in each token it picked. People us

Mathematical statistician working on official statistics in the day and machine learning the rest of the time!

And why to be careful with API LLM response data

What Your LLM API Isn't Telling You About Token Probabilities   

github

Read the articles written by Robert Sneiderman on Peerlist

Articles written by Robert Sneiderman • Peerlist

Machine Learning

computervision

Deep Learning

Pytorch

Flask

Data Science

TensorFlow

research