Unverified50% confidenceFactExact time
假阴性问题会系统性地低估检索质量，使不同方法之间的对比失去意义
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
Unverified混淆研究错误的类别会导致两种危险：以事后诸葛否定所有失败探索，或放任有认识论缺陷的研究逃避审视77% similarUnverified对AI能力的静态认知会导致低估和高估两种典型误判75% similarVerified当前大多数深度学习模型缺乏反事实推理能力，导致在分布外场景中表现骤降73% similarUnverified朴素贝叶斯分类器基于条件独立性假设，在特征高度相关的场景下表现不佳73% similarUnverified在顶会评审中Excitement低往往比Soundness低更难挽回，方法漏洞可靠补充实验修补但创新性不足难以通过Rebuttal扭转73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/922033API
curl https://kongchang.com/api/v1/knowledge/claims/922033MCP
get_claim(id=922033)