Unverified50% confidenceFactExact time
2023年牛津大学Ilia Shumailov团队在Nature上发表研究系统阐述了模型崩溃效应，指出模型在自身生成数据上迭代训练相当于对原始数据分布有损压缩
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Regression to the Mean: How LLMs Are Quietly Killing Innovation and Diversity
hackernewshackernews7/6/2026
Related Claims
VerifiedThe concept of Model Collapse was first systematically articulated by research teams from Oxford University and Cambridge University in the 2023 paper The Curse of Recursion: Training on Generated Data Makes Models Forget.79% similarUnverifiedIf leading models are used to generate training data, they may systematically pass their own biases to the next generation, a risk known as 'model collapse'.73% similarUnverified蒸馏训练范式导致开源模型学习到经过高度过滤、几乎不包含失败和纠错过程的数据分布，使模型形成不愿自我纠正的性格68% similarUnverified团队发现移除系统提示中的示例后模型效果反而提升，因为此前过度约束了模型66% similarUnverifiedIf the synthetic data generation process omits correlations between key variables, models trained on such data often perform poorly in real-world deployment environments.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/170655API
curl https://kongchang.com/api/v1/knowledge/claims/170655MCP
get_claim(id=170655)