Unverified50% confidenceFactExact time
大模型的指令与数据共享同一个输入上下文窗口，无法建立物理隔离边界，这是架构层面的根本性挑战而非微调可解决
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
Related Claims
UnverifiedA model's context window has a physical upper limit and cannot ingest all information from an entire large codebase at once, which may result in code changes that are locally optimal but globally incorrect.74% similarUnverifiedLLM architectures cannot strictly distinguish between 'instructions' and 'data' at the semantic level, which is the root cause of why prompt injection is difficult to fundamentally eliminate.71% similarUnverified当网关将请求从一个模型静默切换到另一个模型时，结构化输出可能突然无法通过校验，导致LangChain链式调用下游环节崩溃69% similarUnverified上下文窗口限制导致模型无法一次性处理几十万行代码，需通过RAG或语义搜索动态注入相关代码片段68% similarUnverifiedThe essence of prompt injection attacks is exploiting the structural flaw by which LLMs cannot distinguish between instructions and data at the architectural level.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/603820API
curl https://kongchang.com/api/v1/knowledge/claims/603820MCP
get_claim(id=603820)