[KongchangAI]
Unverified50% confidenceFactExact time

该研究关注模型在自然日常推理任务中自发出现的不忠实现象(in the wild),而非人为构造的对抗性提示

1
Sources
50%
Confidence
Long-term
Relevance
8/22/2026
First Seen

Sources

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/785641
API
curl https://kongchang.com/api/v1/knowledge/claims/785641
MCP
get_claim(id=785641)