But it can't, we see models get larger and larger and larger models perform better. <Thinking> made such huge improvements, because it makes more text for the language model to process. Cavemanising (lossy compression) the output does it to the input as well.

but some tokens are not really needed? This is probably bad because it is mismatched with training set, but if you trained a model on a dataset removing all prepositions (or whatever caveman speak is), would you have a performance degradation compared to the same model trained on the same dataset without the caveman translation?