Finding GPT-4’s mistakes with GPT-4

Company news OpenAI Source language: English

CriticGPT, a model based on GPT-4, writes critiques of ChatGPT responses to help human trainers spot mistakes during RLHF

Excerpt

CriticGPT, a model based on GPT-4, writes critiques of ChatGPT responses to help human trainers spot mistakes during RLHF

Short excerpt from the source — open the original for the full text.

Sources

← Back to the daily brief