HuggingFace Daily Papers（社区热门论文）· HuggingFace Daily Papers（社区热门论文）·· 9 天前AI 评分42
DISCO：用接地-推理解耦实现分布式长上下文扩展
AI 导读
针对长上下文推理质量随输入增长而崩塌的"context rot"问题，研究者提出 DISCO，借鉴 Apache Spark 的分布式思路，将长上下文切分给多个 Worker LLM 并行做局部接地，由经 GRPO 强化学习训练的 Driver LLM 负责规划与汇总。
正文
Abstract: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 dedicated exclusively to parallel, localized grounding. A central Driver LLM, trained via Reinforcement Learning (GRPO) to optimize planning, orchestrates execution by dynamically mapping queries into atomic extraction tasks and reducing the gathered evidence to synthesize a final answer. By isolating reasoning from raw context noise, DISCO effectively eliminates context rot. On RULER-QA (1M tokens), it maintains 78.4% accuracy where standard baselines collapse. Furthermore, it outperforms full-context models by up to 9.8 points on LongBench v2 and matches frontier models like Gemini-3-Pro-Preview while reducing inference costs by over 80%, establishing a highly efficient paradigm for robust long-context inference.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2609.33485 [cs.CL] |
| (or arXiv:2609.33485v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2609.33485 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Guanzheng Chen [view email]
[v1]
Sun, 27 Sep 2026 11:52:58 UTC (4,564 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org