Smells like cheap text classification. Gonna try this:

...we use gzip with a compressor-
based distance metric to do text classification.
Our method achieves an accuracy comparable to non-pretrained neural network classifiers on in-
distribution datasets and outperforms both pre-trained and non-pretrained models on out-of-
distribution datasets. We also find that our method has greater advantages under few-shot settings.

#nlp #programming
https://aclanthology.org/2023.findings-acl.426/