3 ms·
Seems a lot like Normalized Compression Distance (https://en.wikipedia.org/wiki/Normalized_compression_distance https://en.wikipedia.org/wiki/Normalized_compres
by woliveirajr 4mo ago
Seems a lot like Normalized Compression Distance (https://en.wikipedia.org/wiki/Normalized_compression_distance https://en.wikipedia.org/wiki/Normalized_compression_distanc...) and Kolmogorov Complexity.
- srean 4mo agoGood compression, necessarily, requires good "next token" probability estimation. I was essentially using zlib's implicit probability estimator.