ChatPaper.aiChatPaper

Maglev: Sliding Recurrent Memory

Maglev: Sliding Recurrent Memory

August 5, 2026
Autores: Bo Liu, Qiang Liu
cs.AI

Resumen

We introduce , a recurrent Transformer architecture with fixed-size memory that generalizes sliding-window attention while remaining parallelizable during training. consists of two coupled models: a prefiller Q, which leverages full attentionIn practice, we use interleaved full and sliding-window attention for Q, as this yields stronger performance. The essential requirement is that Q be more expressive than P, with access to the full history. to produce memory targets m'_t, and a decoder P, which uses only sliding-window attention and recurrent K/V injection to produce decoder memories m_t for next-token prediction. We train with a memory consistency loss that aligns m_t with m'_t, allowing inference to use P alone. Empirically, improves validation loss and downstream pretraining benchmarks over sliding-window and latent recurrent transformer baselines. Moreover, sharing parameters between P and Q reduces parameter memory while preserving most of the gains.
English
We introduce , a recurrent Transformer architecture with fixed-size memory that generalizes sliding-window attention while remaining parallelizable during training. consists of two coupled models: a prefiller Q, which leverages full attentionIn practice, we use interleaved full and sliding-window attention for Q, as this yields stronger performance. The essential requirement is that Q be more expressive than P, with access to the full history. to produce memory targets m'_t, and a decoder P, which uses only sliding-window attention and recurrent K/V injection to produce decoder memories m_t for next-token prediction. We train with a memory consistency loss that aligns m_t with m'_t, allowing inference to use P alone. Empirically, improves validation loss and downstream pretraining benchmarks over sliding-window and latent recurrent transformer baselines. Moreover, sharing parameters between P and Q reduces parameter memory while preserving most of the gains.