Unverified50% confidenceFactExact time
Reference model-free attacks achieve, in most scenarios, results statistically comparable to those of attacks that rely on adversarial knowledge.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
Unverified数据投毒攻击按攻击目标可分为可用性攻击和目标攻击，前者旨在降低模型整体性能，后者只在特定触发条件下改变模型行为74% similarUnverifiedIndirect prompt injection specifically refers to attackers carrying out attacks through external content that the model actively retrieves, making it more dangerous.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71% similarUnverified工具调用使攻击面从让模型说错话升级为让模型做错事，危害从信息误导上升为实质性的数据访问和系统篡改71% similarUnverifiedAny scenario in which instructions and untrusted data share the same context (such as prompt injection, RAG, or agents) carries the risk that the model may follow information it was told to ignore.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1023954API
curl https://kongchang.com/api/v1/knowledge/claims/1023954MCP
get_claim(id=1023954)