Unverified60% confidenceFactExact time
In 2023, researchers demonstrated that contaminating less than 1% of training data is sufficient to alter model behavior under specific trigger conditions
2
Sources
60%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
以色列斥资4650万美元干预AI舆论：生成式AI成为信息战新战场
Related Claims
UnverifiedIn 2023, researchers from Oxford University and other institutions published a study finding that after only 5–10 generations of recursive training, the degradation in model output quality became visible to the naked eye.77% similarUnverifiedResearch indicates that an attacker needs to modify as little as 0.01% of training data to potentially implant backdoor behaviors in a model that are undetectable by standard evaluations.74% similarUnverifiedAnthropic's research shows that even mild alignment training can cause a 2–8 percentage point decline in model performance on certain reasoning-intensive tasks.73% similarUnverified2024年研究显示，经过专门训练的小型改写模型可在保持95%以上语义相似度的同时将水印检测率降至接近随机水平72% similarUnverifiedExperiments conducted by the research team on models such as GPT-2 showed that after only 5–9 generations of recursive training, the quality of model outputs degrades significantly.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/699344API
curl https://kongchang.com/api/v1/knowledge/claims/699344MCP
get_claim(id=699344)