Any clever strategies for coherence when using long contexts with gpt-3.5-turbo-16k? Seems to still have trouble focusing (unlike 4).

→ View original post on X — @localghost