---
id: 20260606-T0-04
title: "模型崩溃新发现：合成数据污染通过双层SIR传播"
title_en: "Model Collapse Spread via Bilayer SIR Dynamics"
url: https://ai.daily.yangsir.net/daily/20260606-T0-04
issue_date: 2026-06-06
publish_date: 2026-06-05T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2606.05168
---

# 模型崩溃新发现：合成数据污染通过双层SIR传播

arXiv论文揭示合成数据导致模型崩溃的传播机制。传统观点认为崩溃是单链退化，但实际中模型间存在交叉污染：一个模型产生合成数据，被其他模型吸收，又生成新文本。这种双层SIR动态机制加速了整个AI系统的知识退化。

## English Version

**Model Collapse Spread via Bilayer SIR Dynamics**

arXiv paper reveals how synthetic data contamination spreads model collapse. Traditionally seen as single-chain degradation, real-world systems have cross-contamination: one model generates synthetic data absorbed by others, creating new text. This bilayer SIR dynamic accelerates knowledge degradation across AI systems.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2606.05168)

**详情页**：https://ai.daily.yangsir.net/daily/20260606-T0-04

---

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