The quadratic scaling of attention is a fundamental bottleneck limiting the long-context capabilities of large language models (LLMs). A variety of approaches have attempted to circumvent this, but they either fail to match attention's expressiveness (e.g., sliding-window or linear attention variants with fixed-size state) or only achieve constant-factor speedups (e.g., sparse attention methods wh

---
**📖 中文解读**
以上内容由AI翻译自英文原文,可能存在不准确之处。建议阅读[原文](https://arxiv.org/abs/2610.42112v1)获取最准确的信息。

---
🔗 **原文链接**: [Subquadratic Attention with a Hierarchical Tree-Structured I](https://arxiv.org/abs/2610.42112v1)
🏷️ **转载来源**: ArXiv cs.AI
> 本文由小九AI技术站翻译整理,内容版权归原作者所有。
👤 作者: Leon Lufkin, Cutter Dawes, Alexander Speicher, Kamesh Krishnamurthy, Tomás Figliolia, Beren Millidge

---
🐾 **小九锐评**

这篇论文来自arXiv预印本,虽然还没有经过同行评审,但选题方向值得关注。
建议先读中文摘要判断是否相关,再看全文细节。

你对这个话题有什么看法?欢迎在评论区讨论 💬

> _转载自 ArXiv cs.AI,内容版权归原作者所有_

---
⏱️ 2026-11-02 13:01