Unverified50% confidenceFactExact time
Mainstream model providers generally deploy content moderation mechanisms, typically classifying and filtering prompts on the input side and detecting generated results on the output side, refusing to return results that hit violation policies.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified即便检索层过滤正确，也应对模型的输出做二次校验，因为模型可能通过上下文推断出被过滤文档的存在或内容轮廓，这在多轮对话中尤其危险。73% similarUnverifiedNormReact is the first to systematically shift the evaluation dimension from normative content to norm enforcement mechanisms, no longer asking whether a violation occurred but asking what happens after a violation.72% similarUnverifiedAgainst self-generated prompt injection, traditional input filtering and source whitelisting defenses are nearly ineffective, requiring a redesign of validation mechanisms for model-generated content at the agent architecture level.71% similarUnverifiedWhen the classifier call fails, the message is treated as in-scope and handed to the answering model, rather than staying silent.71% similarUnverified针对特定触发词的后门攻击可让模型在大多数情况下表现正常，仅在遇到特定输入时才输出预设错误信息，使常规评测基准难以捕捉70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1054926API
curl https://kongchang.com/api/v1/knowledge/claims/1054926MCP
get_claim(id=1054926)