Novel Memory Forgetting Techniques for Autonomous AI Agents: Balancing Relevance and Efficiency

This paper introduces adaptive budgeted forgetting techniques for autonomous AI agents. It addresses the degradation of long-running AI agents due to uncontrolled memory growth, proposing a three-dimensional scoring system (recency, frequency, semantic alignment) to manage memory. The research shows improved long-horizon F1 scores and reduced false memories, leading to faster boot times and cleaner context for AI agents.