Unverified50% confidenceFactExact time
A user study with 10 human annotators shows that the consistency of LLMaJ judges under synonymous rephrasing drops 6-9 times as much as that of humans.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn document deskewing, even a tilt of 3–5 degrees can cause recognition rates to drop by more than 10%.72% similarUnverifiedThe diarization error rate (DER) on standard evaluation sets has dropped below 10%.71% similarUnverifiedWhen Humanbound cannot obtain a score, it fabricates a score of 5/10, which may cause test reports to contain false data.71% similarUnverified在事实性评估任务上模型评判者准确率仍显著低于人类专家，但在风格、连贯性等主观维度已接近人类评审员间一致性水平68% similarUnverifiedIn 11 of the 12 models, between 87.5% and 100% of capitulation cases stopped at the exact wrong number provided by the user, meaning capitulation typically involves switching sides entirely.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114681API
curl https://kongchang.com/api/v1/knowledge/claims/1114681MCP
get_claim(id=1114681)