Unverified50% confidenceTradeoffExact time
When semantically similar but inaccurate content is retrieved, the model may output incorrect recollections with high confidence, with no warning mechanism at the system level — this is the false memory problem.
1
Sources
50%
Confidence
Long-term
Relevance
10/3/2026
First Seen
Sources
Related Entities
Related Claims
Unverified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79% 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78% 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77% similarUnverifiedIncorrect memory retrieval can be worse than having no memory at all.77% similarUnverifiedLarge language models may generate content that appears credible but is completely wrong (hallucinations), thereby misleading users' decisions.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/970951API
curl https://kongchang.com/api/v1/knowledge/claims/970951MCP
get_claim(id=970951)