Do Language Models Need Sleep? Offline Recurrence for Improved Online Inference

Sleep health notes

Sleep health notes will be added by the Sleepatch editorial team.

中文解读

中文解读待补充：本站会优先为失眠研究、睡眠质量改善、昼夜节律等高价值睡眠研究添加中文说明。

Original abstract

Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs $N$ offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving the latency of wake-time prediction. We test our method on controlled synthetic tasks, including cellular automata and multi-hop graph retrieval, as well as a realistic math reasoning task, on which a regular transformer as well as SSM-attention hybrid models fail. We then show that increasing sleep duration $N$ for our models improves performance, with the largest gains on examples that require deeper reasoning.

Links and sources

• Official / arXiv page
• PDF from original source

Comments