HN Explorer

Ask HN: How does everyone keep track of user reactions to their LLM output?

2 points by vsroy ↗ HN
For example: If you make a chat bot and let the user regenerate their responses; then there is an implicit preference ranking of the chatbot responses. Do any of you keep track of this?

1 comment

[ 3.6 ms ] story [ 12.2 ms ] thread
ksaj
On ChatGPT you can give feedback. Also there is a sort of upvote checkmark thing that I think helps them gauge good response patterns to feed back into their updated versions.