Unverified50% confidenceBenchmarkExact time
In the author's tests, a threshold of 100 caused about 14% of batches to lose small features such as whisker lines and internal detail lines.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThis benchmark uses 10 original short documents of 200 to 250 words each, with 18 planted anomalies, 37 scoring entries in total (18 dirty and 19 clean), testing 13 models with 74 prompts per model.67% similarUnverifiedResearch from institutions including Stanford University shows that for short texts under 200 words, the false positive rate of existing detectors can be as high as tens of percentage points.65% similarUnverifiedAmong the five evaluators the authors tested, if the answer key had minor errors, the score was only 0.21 — lower than giving no key at all (0.46) and lower than giving a completely irrelevant key (0.50); while a human-verified model-generated answer scored 0.61, close to the 0.69 of a human-written one.65% similarUnverifiedFor a document containing 100 fields, even if the per-field accuracy is as high as 99%, the probability of the entire document being fully correct is only about 37%.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1043122API
curl https://kongchang.com/api/v1/knowledge/claims/1043122MCP
get_claim(id=1043122)