Unverified60% confidenceFactExact time
A language model never calls tools directly; it only outputs a piece of text stating which tool it wants to call, after which code executes the call and returns the result as text.
2
Sources
60%
Confidence
Long-term
Relevance
9/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified模型调用工具时会将输出从普通文本切换为结构化的函数调用格式（finish_reason: tool_calls）77% similarUnverified提示词工程无需修改模型权重，仅凭文本指令即可实现任务切换74% similarUnverifiedLarge language models cannot structurally distinguish between user instructions and instructions embedded in content, because both are part of the same token sequence from the model's perspective.71% similarUnverified提示词注入分为直接注入和间接注入两种形式，因语言模型无法区分数据与指令，凡进入上下文窗口的文本都可能被执行71% similarUnverifiedA model is essentially just a function: text in, text out, with no ability to search the web, read files, run code, or remember the previous message.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/953110API
curl https://kongchang.com/api/v1/knowledge/claims/953110MCP
get_claim(id=953110)