News for hackers. Stuff that matters.

A comment on GPT-4 LLM simulates people well enough to replicate social science experiments.

Re:GPT-4 LLM simulates people well enough to replicate socia... by dartos on Wednesday August 07, 2024 @08:33PM (#41186932)
Were those experiments in the training set?

If so, how close was the examination vs the record the model was trained on.

Some interesting insights there, I think.

[ Parent | on HN ]
Re:GPT-4 LLM simulates people well enough to replicate socia... by masterofpupp3ts on Wednesday August 07, 2024 @08:55PM (#41187044)
The answers to your questions are in the paper linked in the first line of the app
[ Parent | on HN ]
Re:GPT-4 LLM simulates people well enough to replicate socia... by cpeterso on Thursday August 08, 2024 @12:08AM (#41188012)
> Accuracy remained high for unpublished studies that could not appear in the model's training data (r = 0.90).
[ Parent | on HN ]