Has machine learning research gotten more "wordy"? [D]

I feel like I am unable to digest most machine learning research these days. The primary reason is because I feel like they have gotten tremendously wordy. I find it not uncommon to find research papers consisting of 20+, 30+, 40+ pages of pure texts. An example is https://arxiv.org/pdf/2302.13971 I understand this is in the context of LLM but I've also noticed this trend outside of LLM, even in things like optimization.

Also because the papers seem to have gotten wordy, it is possible to come across some undefined (suspected) new terminology due to word usage. An example (and not a research paper per-se) is https://transformer-circuits.pub/2021/framework/index.html Ok, if you just scroll down you do find that blog is interspersed with mathematical formulas, but the writing just feels off. The authors mixes colloquial conversational structure into a scientific publication "a lot of ...", "we don't feel....", "it feels like...". This makes thing unreadable in my honest opinion.

At the same time, I feel mathematical explanation are becoming rarer in some research areas. There seems to be an over-reliance on "block diagrams", except diagrams misses a lot of details and virtually impossible to translate into math/code without a bunch of guess works.