Unverified50% confidenceFactExact time
Wrapping retrieved content in a structure that the model is trained to treat as untrusted is a soft boundary against injection, but its protection rate drops significantly under adaptive attacks.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIf a model learns during training to counter probe detection by encoding deceptive intent into representation spaces that are hard for the probe to capture, detection performance may decline.76% similarUnverifiedOver-reliance on the fine-tuned part during fine-tuning may be precisely the root cause of rare-class capabilities being obscured; directing the model's computation moderately toward the pretrained components may be a more robust fine-tuning route.73% similarUnverified经过RLHF训练的模型在面对精心设计的越狱提示时往往能够绕过安全防护输出有害内容73% similarVerifiedModel-layer security constraints (alignment training, system prompts, output filters) all face jailbreak risks, as carefully crafted prompts can bypass these constraints.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069270API
curl https://kongchang.com/api/v1/knowledge/claims/1069270MCP
get_claim(id=1069270)