flâneur — a map of the web's best reading

Learning to Summarize with Human Feedback

openai.com · saved by 1 readers

We've applied reinforcement learning from human feedback to train language models that are better at summarization. Our models generate summaries that are better than summaries from 10x larger models trained only with supervised learning. Even though we train our models on the Reddit TL;DR dataset, the same models transfer

Explore this link on the map →