DISCO: Distributed Long Context Scaling with Grounding-Reasoning Disaggregation

  • 类型：arxiv
  • 标识：2609.33485
  • 链接：https://arxiv.org/abs/2609.33485
  • 主分类：llm-infra
  • 形态：method
  • TLDR：While Large Language Models (LLMs) advertise million-token context windows, reasoning quality often collapses as inputs grow -- a phenomenon termed context rot. This failure stems from a structural entanglement in monolithic architectures, where the massive search burden of contextual grounding exhausts the representational capacity needed for complex reasoning. To resolve this, we propose Grounding-Reasoning Disaggregation via DIStributed long COntext scaling (DISCO). Inspired by distributed computing frameworks like Apache Spark, DISCO partitions long context across a fleet of Worker LLMs de
  • 待LLM分类：否
  • 标题中文：DISCO: 基于接地-推理解耦的分布式长上下文扩展
  • TLDR中文：尽管大语言模型（LLM）宣称支持百万 token 上下文窗口，推理质量却常随输入增长而崩溃——即"上下文腐烂"。该失败源于单体架构中的结构性纠缠：上下文接地的巨大搜索负担耗尽了复杂推理所需的表征能力。我们提出基于接地-推理解耦的分布式长上下文扩展（DISCO），受 Apache Spark 等分布式计算框架启发，将长上下文切分到一组 Worker LLM 上进行
  • 成熟度：research
  • 场景：长上下文、分布式推理、注意力优化
  • 来源文件：
  • /inbox/tom/_candidates/2026-09-29-agent-rag-longcontext-candidates.json