Unverified50% confidenceCautionExact time
If the data always provides tools that get used, the model learns the erroneous association that it must use a tool whenever it sees one.
1
Sources
50%
Confidence
Long-term
Relevance
9/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified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76% similarUnverifiedError messages should be written for the model to read and must be "actionable" rather than "technically correct" — throwing raw parser errors at the model is of no value.74% similarUnverifiedInference methods based on model internal features face a trade-off between false positives and false negatives72% similarUnverified系统会试探性地应用补丁，若失败则将错误信息反馈给模型进行修正，构成闭环纠错机制70% similarUnverifiedIn Agentic scenarios, sources of error may include prompt design flaws, tool API return formats not matching expectations, information loss during multi-step reasoning, or gaps in the model's domain knowledge.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/928824API
curl https://kongchang.com/api/v1/knowledge/claims/928824MCP
get_claim(id=928824)