Why Do Models Look Worse on Enterprise-Length Docs Than Short Docs?

https://super-wiki.win/index.php/DeepSeek-R1_shows_14.3%25_and_11.3%25_on_Vectara%E2%80%94is_it_usable_for_summaries%3F

In my eleven years of working in applied NLP, I have seen a recurring cycle of hype followed by "enterprise disillusionment