Unverified50% confidenceFactExact time
战士「Imp」仅用一条指令不断将自身复制到下一个内存地址，「Dwarf」采用定期轰炸策略每隔固定间隔向内存写入DAT指令
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
UnverifiedSGLang通过RadixAttention实现基于哈希的前缀缓存机制，命中缓存则直接复用已有KV张量并跳过prefill阶段重复计算56% similarUnverifiedMemPalace periodically reminds the AI to save key memories through a Hooks mechanism and triggers an emergency save before the context is about to be compressed.56% similarUnverifiedThe goal of memory poisoning attacks is to embed malicious instructions in an Agent's persistent memory so that it executes unintended operations in future interactions, which differs from the immediate hijacking of prompt injection.55% similarUnverified标准的LLM调用每次都是独立的，无法记住上下文，因此Agent需要记忆系统解决无状态缺陷54% similarUnverified越狱攻击经历了从简单角色扮演（如DAN提示）到Base64编码绕过、多轮对话渐进式诱导（Crescendo Attack）、对抗性后缀攻击等多个演进阶段54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/529735API
curl https://kongchang.com/api/v1/knowledge/claims/529735MCP
get_claim(id=529735)