Unverified50% confidenceFactExact time
System prompt adherence is not a hard constraint but a probabilistic tendency; models may ignore system instructions when user input creates a semantic conflict with the system prompt.
1
Sources
50%
Confidence
Long-term
Relevance
8/15/2026
First Seen
Sources
Google AI模型突然变笨？模型退化现象的成因与应对策略
Related Claims
UnverifiedBecause large language models are based on probabilistic sampling, even if instructions are written in the context, the model may still forget or skip certain instructions in complex reasoning chains, which is known as instruction-following instability.77% similarUnverified系统提示词是一种零样本引导方式，在推理阶段调控模型行为，不同于微调会改变模型的权重参数75% similarUnverifiedA well-aligned model will guide meta-prompts toward compliant creative expression, while an insufficiently aligned model may exhibit unexpected behavior.74% similarUnverifiedCommon system prompt extraction techniques include directly asking the model to output its system prompt, using role-playing to bypass restrictions, and exploiting the model's tendency toward excessive compliance for indirect elicitation.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/751775API
curl https://kongchang.com/api/v1/knowledge/claims/751775MCP
get_claim(id=751775)