Maglev: Sliding Recurrent Memory

• 类型：arxiv
• 标识：2608.02870
• 链接：https://arxiv.org/abs/2608.02870
• 主分类：engineering
• 形态：method
• 被引：0
• 被引来源：Semantic Scholar
• S2被引：0
• 影响力被引：0
• TLDR：A recurrent Transformer architecture with fixed-size memory that generalizes sliding-window attention while remaining parallelizable during training and improves validation loss and downstream pretraining benchmarks over sliding-window and latent recurrent transformer baselines.
• 待LLM分类：否
• 标题中文：Maglev：滑动循环记忆
• TLDR中文：一种具有固定大小记忆的循环 Transformer 架构，可泛化 sliding-window attention，同时在训练期间保持并行性，并在验证 loss 和下游预训练 benchmark 上优于 sliding-window 和 latent recurrent Transformer 基线。
• 来源文件：
• /inbox/tom/_candidates/2026-08-15-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-16-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-17-agent-rag-longcontext-candidates.json
• /inbox/tom/_candidates/2026-08-17-agent-memory-tool-use-candidates.json
• [S2 enrich]