[sumtechguy]

Good summary.

The LLM vs a static tree has some interesting oppositions. With a static tree as emitted by a compression alg will probably many times beat an LLM. As it has full knowledge of the whole stream (or in gzips version that window). So it can do things where it can look back and say 'hm the tree I spit out was not that good let me build a better one'. Where as an LLM does not really have that before hand knowledge. Using a pre-cooked LZW tree for all inputs would be more akin to using an LLM.

[willvarfar]

I would envisage the LLM is allowed to train on each and every input token. So, to begin with, it knows nothing; but to predict the very last token, it has internalised the whole preceding stream.

Now I wouldn't expect it to be particularly competitive in enwik8 or enwik9, but the question would be: is there any max-model-size and input-length for which it would right now pull ahead and become the best known or at least competitive predictor?

[sumtechguy]

I would expect it to be 'ok'. Basically as if you used a pre-trained LZW table only and shipped that along with each stream but the results would be mixed. A compressor has the advantage of foresight and hindsight whereas a LLM would only have hindsight. As any input stream would be basically at the mercy of the previous streams fed into it. Those may or may not be optimal.

It is an interesting hypothesis. But my gut feeling is I would expect a LLM to perform on average worse. Competitive? Yes, but still worse. But it is something I am sure someone will test.

From the hundreds of different compressor models I have made for myself over the years. Usually believe it or not the compressed data is usually the best part. It is the decode tree/table/key/whatever that usually ends up crowding out the savings on the compressed data. In this case it would be the LLM weights or whatever the LLM spits out for the tree/decode.