Tag
RecMem is a recurrence-based memory consolidation method for long-running LLM agents that reduces token consumption by up to 87% while improving accuracy, by only invoking LLMs when semantically similar interactions recur.