Memory Decay
A phenomenon in which AI agents performing long-context tasks lose the ability to track their own state and progress because historical information is truncated or compressed.
Timeline (last 90 days)
Memory decay refers to the truncation or compression of historical information once it exceeds the context window, significantly reducing an agent's ability to track its own inventory, completed progress, and outstanding commitments.
Memory decay has three main engineering implementations: time-based explicit expiration (TTL), usage-frequency-based adaptive decay, and reliance on implicit feedback signals.
All Facts (2)
Memory decay refers to the truncation or compression of historical information once it exceeds the context window, significantly reducing an agent's ability to track its own inventory, completed progress, and outstanding commitments.
50%UnverifiedMemory decay has three main engineering implementations: time-based explicit expiration (TTL), usage-frequency-based adaptive decay, and reliance on implicit feedback signals.
50%