search

zjysteven / mink-plus-plus

Min-K%++: Improved baseline for detecting pre-training data of LLMs https://arxiv.org/abs/2404.02936
https://zjysteven.github.io/mink-plus-plus/
MIT License
26 stars 5 forks source link

Training code #3

Closed codeboy5 closed 3 months ago

codeboy5 commented 3 months ago

Hey, would like to reproduce some of the results from the paper. Is the code to train the models available somewhere?

zjysteven commented 3 months ago

Hi, not sure if I followed your question closely. The proposed method (and all considered baselines) don't involve any training. What exact training are you referring to?

zjysteven commented 3 months ago

Closing due to inactivity.