Verified80% confidenceFactExact time
Transformer-based large language models use attention mechanisms to uniformly encode system prompts, user inputs, and externally retrieved content as homogenized token sequences, making it impossible for the model to distinguish trusted instructions from untrusted data at the semantic level.
8
Sources
80%
Confidence
Long-term
Relevance
7/20/2026
First Seen
Sources
Related Claims
VerifiedThe memory of large language models fundamentally relies on the attention mechanism of the Transformer architecture, and its effective information capacity is limited by the number of tokens in the context window.81% similarUnverified提示词注入的核心原理在于LLM无法从根本上区分系统指令和用户输入，因为Transformer架构的注意力机制对所有输入token一视同仁，系统提示词与用户输入的token在数学上处于同一向量空间78% similarUnverifiedTransformer-based large language models generate text token by token in an autoregressive manner, with no external symbolic verification system to check whether each reasoning step is valid77% similarUnverifiedCurrent mainstream large language models generally adopt the Transformer architecture, whose self-attention mechanism has a computational complexity that grows quadratically with sequence length (O(n²)) when processing long contexts.77% similarUnverifiedThe Transformer architecture relies on key tokens in context to build semantic representations, and removing low-information function words such as articles typically does not significantly impair the model's grasp of core semantics.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/575582API
curl https://kongchang.com/api/v1/knowledge/claims/575582MCP
get_claim(id=575582)