Unverified70% confidenceFactExact time
Aligned models tend to proactively introduce disclaimers or interrupt conversation flow when handling hypothetical reasoning and role-playing tasks, behaviors that are almost nonexistent in base models.
2
Sources
70%
Confidence
Long-term
Relevance
7/25/2026
First Seen
Sources
Gemini 3.0 Flash比3.5更受欢迎？揭秘AI「对齐税」真相
redditr/Bard7/12/2026
Related Claims
UnverifiedWhen constraining output format through prompts, the model may forget these constraints in long conversations or complex reasoning scenarios.80% similarUnverifiedThe thin adapter layer strategy keeps core task logic unchanged and only adjusts the interface portion that interacts with the model, designing pre-prompts or post-processing logic tailored to the reasoning characteristics of different models.78% similarVerifiedThe reasoning retention feature only takes effect on models that support reasoning output and does not affect regular conversational models.77% similarVerifiedEach inference call to a large model is an independent forward pass; the model weights are completely fixed during inference and there is no mechanism to preserve state across conversation sessions.77% similarVerifiedA well-aligned model will guide meta-prompts toward compliant creative expression, while an insufficiently aligned model may exhibit unexpected behavior.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/611256API
curl https://kongchang.com/api/v1/knowledge/claims/611256MCP
get_claim(id=611256)