Unverified50% confidenceOpinionExact time
Some commenters argue that having fake data cited by LLMs is a natural extension of SEO black-hat tactics into the generative era, with the goal shifting from search ranking to model citation.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
UnverifiedLLM生成的虚假学术引用往往具有极高的可信外观，作者名、期刊名、年份、标题都遵循合理的分布，但组合结果可能从未真实存在65% similarUnverifiedThe risk of pseudo-labels lies in confirmation bias—erroneous predictions get reinforced; therefore in practice novelty detection (e.g., Mahalanobis distance) is usually combined to judge whether the distribution has drifted, and pseudo-label updates are enabled only when the drift is significant and prediction confidence is high.62% similarUnverifiedTraditional pseudo-labeling suffers from confirmation bias: repeatedly using pseudo-labels from the same model causes biases to accumulate and amplify with each iteration.62% similarUnverifiedLLM自动评估检索结果时可能存在偏差，如偏好长度更长、语言更流畅的段落，或对特定领域知识有盲区，因此小规模人工标注金标准仍不可替代。62% similarUnverifiedCitation decay monitoring needs to distinguish two phenomena: first, retrieval-layer loss, i.e. the page is no longer included in the retriever's candidate set; second, generation-layer replacement, i.e. the page is still retrieved, but the model chose other sources as citations when synthesizing the answer.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1075905API
curl https://kongchang.com/api/v1/knowledge/claims/1075905MCP
get_claim(id=1075905)