1 paper · 1 filter
Rishabh Sharma, Rishika Lall
Does conversational memory need LLM-extracted facts, or is selecting the right raw turns enough? Published results disagree. Extraction-based systems report gains from distilled fa…