Do agents need to sleep?

The paper argues that large language model agents, which persist across sessions and accumulate memory, suffer from unchecked plasticity leading to semantic drift, retrieval errors, and memory saturation. It draws on a selectionist theory of sleep—comprising ringing, resonance, and renormalization—to propose that each phase can be implemented in digital memory systems. The authors predict that selectionist consolidation will outperform instructionist methods as memory stores grow, suggesting that sleep is a necessary cost of continuous learning under finite capacity.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

• llms
• agents