Unverified90% confidenceFactTime unknown

合成数据的反馈循环会导致模型崩溃，表现为词汇单一化、句子结构趋同、长尾内容消失

1

Sources

90%

Confidence

Long-term

Relevance

5/31/2026

First Seen

Sources

Synthetic Data: Cure or Poison? Breaking Through the AI Training Data Drought

bilibili老傅1024

Related Entities

Model Collapseconcept

Related Claims

Unverified端到端模型改变的主要是架构，而许多失败模式（噪声、口音、语码转换）的根源在于数据分布而非架构69% similar Unverified标准自回归模型逐Token从左到右生成文本，在时间步t仅依赖之前的输出，缺乏自我审查和回溯机制，导致错误积累效应69% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/9800

API

curl https://kongchang.com/api/v1/knowledge/claims/9800

MCP

get_claim(id=9800)