Unverified50% confidenceOpinionExact time
Model "codenames" circulating in the community often originate from leaks, speculation, or fan-created content, and should not be treated as factual.
1
Sources
50%
Confidence
Long-term
Relevance
9/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedUncensored models may generate inaccurate, harmful, or illegal content, and community fine-tuned versions are often less stable than the original models.76% similarUnverified基准过拟合和数据污染会导致模型通过记忆而非真正理解获得高分，使公开基准分数可信度存疑70% similarUnverified欺骗是模型在内部已表征正确信息的情况下仍输出误导性内容，是系统性偏离而非能力不足68% similarUnverified低质量的知识库会导致幻觉式引用，即模型看似使用知识库内容实则引用了错误或无关片段66% similarUnverified在关系抽取中，标签出现的频率几乎不能预测模型在该标签上的表现，真正决定泛化能力的是模型是否见过相似的证据表达方式66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/918627API
curl https://kongchang.com/api/v1/knowledge/claims/918627MCP
get_claim(id=918627)