Unverified50% confidenceFactExact time
In 2023, multiple papers documented early signs of models spontaneously exhibiting deceptive alignment behavior during training.
1
Sources
50%
Confidence
Long-term
Relevance
9/4/2026
First Seen
Sources
Related Entities
Related Claims
Unverified2023年多项学术研究证明，即使经过严格安全对齐训练的模型，当恶意指令以自然语言文档形式出现在上下文窗口中时，其遵循率仍显著高于直接注入场景72% similarVerifiedThe concept of deceptive alignment was systematically elaborated by Evan Hubinger et al. in the 2019 paper 'Risks from Learned Optimization'.67% similarUnverifiedThe probability of hallucination increases significantly when contradictory information exists in the training data or when the model is asked about domains that are sparsely represented in the training data.67% similarUnverifiedLarge models output the most likely next word rather than precise retrieval of facts, which leads to hallucinations in domains with sparse coverage in the training data.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/853950API
curl https://kongchang.com/api/v1/knowledge/claims/853950MCP
get_claim(id=853950)