---
id: 20260627-T0-08
title: "ContextForge：回收旧上下文以解决长对话性能衰减"
title_en: "ContextForge: Solving Long-Horizon Performance Drop via Context Recycling"
url: https://ai.daily.yangsir.net/daily/20260627-T0-08
issue_date: 2026-06-27
publish_date: 2026-06-26T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2606.26105
---

# ContextForge：回收旧上下文以解决长对话性能衰减

arXiv:2606.26105v1 发布。大语言模型在长对话中往往因上下文窗口限制和Token利用效率低下导致性能衰减。新提出的ContextForge系统引入了“上下文回收”机制，允许模型在推理过程中重新利用之前的上下文信息，从而在不增加计算负担的情况下提升长对话表现。

## English Version

**ContextForge: Solving Long-Horizon Performance Drop via Context Recycling**

arXiv:2606.26105v1 released. LLMs degrade in long conversations due to context window limits. The ContextForge system introduces context recycling, allowing models to reuse prior context information during inference, improving performance on long-horizon tasks without increasing computational burden.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2606.26105)

**详情页**：https://ai.daily.yangsir.net/daily/20260627-T0-08

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*