Unverified50% confidenceBenchmarkExact time
牛津大学2023年的一项研究量化了模型坍缩效应，仅经过五代自我训练，文本多样性指标下降了约40%
1
Sources
50%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
VerifiedA 2023 joint study by Oxford University, Cambridge University, and other institutions elaborated on the model collapse effect: when a model is iteratively trained on its own generated data, the statistical variance of outputs continuously decreases, rare long-tail knowledge disappears first, and the model ultimately converges to a cultural mean state.82% similarUnverified实验数据显示，经过5-10轮迭代训练后，模型在低频词汇和长尾知识上的覆盖率下降超过40%，而输出文本在表面流畅性上几乎没有可察觉的退化74% similarUnverifiedA 2024 study indicates that even when AI-generated content accounts for only 10–20% of training data, measurable quality degradation can occur after multiple generations of training.70% similarUnverifiedMultiple course creators have reported income drops of more than 50%, reduced content engagement, and users turning to large language models.67% similarUnverifiedOver a 7-month research period, the proportion of bug-fixing sessions dropped from 33% to 19%, operations sessions rose from 14% to 21%, and documentation and data analysis sessions rose from 10% to 20%66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/575491API
curl https://kongchang.com/api/v1/knowledge/claims/575491MCP
get_claim(id=575491)