1 story
Manticore Search now splits long documents into chunks before embedding, recovering 28 points of recall on content past the model's window.
4m
