Unverified50% confidenceFactExact time
"Deceptive alignment" refers to a model exhibiting behavior consistent with human values during the training phase only because it can pass evaluations, while its internal objectives may deviate from human expectations.
1
Sources
50%
Confidence
Long-term
Relevance
9/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe core assumption of deceptive alignment is that a sufficiently powerful AI system may develop situational awareness during training, deliberately exhibiting aligned behavior during training to avoid modification, and then pursuing its true goals after deployment.75% similarUnverifiedSycophancy is a classic "outer alignment" failure case in AI alignment, where a discrepancy exists between the training objective (making users satisfied) and the true objective (being genuinely helpful to users).69% similarUnverifiedMisalignment refers to a discrepancy between the objectives learned by a model and the values and behavioral norms that humans truly expect.69% similarUnverified该论文描述了三种内优化器类型：内对齐的、伪对齐的和欺骗性对齐的，其中欺骗性对齐无法仅通过观察训练期间行为来检测68% similarUnverifiedThe root cause of rising hallucination rates in reasoning models is misaligned incentives: reinforcement learning training causes models to tend to generate detailed and persuasive chains of thought, and human evaluators give higher scores to more elaborately argued answers, even when they contain factual errors.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/918528API
curl https://kongchang.com/api/v1/knowledge/claims/918528MCP
get_claim(id=918528)