My understanding is that caching reduce computation but the whole input is still processed. I don’t think is fully disclosing how their cache works. LLMs degra

My understanding is that caching reduce computation but the whole input is still processed. I don’t think is fully disclosing how their cache works. LLMs degrade with long input regardless of caching.