Unverified50% confidenceFactExact time
When the prompt format implies "this is an article that needs summarizing," the model activates relevant world-knowledge priors based on entity names (such as person names) in the title, thereby generating content that looks coherent but has no source basis.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedGenerative summarization may introduce information not present in the source text — known as factual drift — because models reorganize content in their own language.67% similarUnverifiedThe extraction quality of keywords and named entities will directly determine the quality of the hypergraph structure.67% similarUnverifiedThe embedding model only sees the sentence pattern of the paragraph and cannot perceive its heading; this is not a flaw of the embedding model, but rather that any fixed-length vector compression discards some information—the problem is that structural clues already provided by the author are being actively discarded65% similarUnverifiedThe essence of hashtags is a decentralized content classification system — it does not require the platform to preset a classification framework, but is created and used bottom-up by users65% similarUnverifiedA mention is merely the brand name or product name appearing in the generated text, possibly from memory in the model's training data, a summary of other sources, or model hallucination, and often no corresponding page can be found.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1027709API
curl https://kongchang.com/api/v1/knowledge/claims/1027709MCP
get_claim(id=1027709)