Unverified50% confidenceOpinionExact time
模型在未收到明确指令时输出图片，意味着系统提示词设计或RLHF训练中存在激励偏差
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
Related Claims
Unverified提示词与模型版本存在深度绑定关系，不同版本模型在RLHF训练数据、指令微调策略上的差异会导致相同提示词产生不同输出，这种现象被称为「提示词漂移」82% similarUnverified大语言模型的指令跟随能力是通过RLHF训练习得的统计倾向，注意力机制无法从架构层面区分硬规则与软建议80% similarUnverified提示注入利用的是模型无法可靠区分「数据」与「指令」的天然缺陷，这一缺陷根植于LLM的训练机制本身80% similarUnverifiedRLHF的人类介入发生在训练阶段并固化为模型权重，属于隐式编码，而表征工程干预在推理时操控内部表征，具有即时性和可逆性79% similarUnverifiedSome systems embed safety preferences into model weights during the training phase through RLHF (Reinforcement Learning from Human Feedback), causing the model itself to tend to refuse generating certain content rather than relying on external filtering.79% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/464857API
curl https://kongchang.com/api/v1/knowledge/claims/464857MCP
get_claim(id=464857)