Unverified50% confidenceConceptExact time
Hallucination metrics differ in definition: the overall proportion of unsupported statements and the "fabricating instead of admitting it doesn't know" ratio are metrics of different dimensions, and the two can simultaneously show a combination of "overall improvement, boundary behavior worsening".
1
Sources
50%
Confidence
Long-term
Relevance
10/29/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedResearch indicates that hallucination rates do not have a linear negative correlation with model scale, and larger models sometimes produce harder-to-detect misinformation in a more fluent manner.79% similarUnverifiedResearch shows that hallucination rate is negatively correlated with the knowledge density of the prompt; the richer and more precise the context provided in the prompt, the lower the probability that the model will generate fabricated content.77% similarUnverified模型规模越大，幻觉的表述往往越流畅、越具迷惑性，识别难度反而上升75% similarUnverifiedHallucinated proofs differ from rounding errors in numerical computation or obvious error reports in symbolic computation, as the latter two tend to be exposed in subsequent steps, whereas hallucinated proofs may remain undetected until discovered by peer review.74% similarUnverifiedLowering the hallucination rate often comes with a higher refusal rate or more conservative answers, so real-world deployment requires a trade-off between accuracy and coverage.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1112335API
curl https://kongchang.com/api/v1/knowledge/claims/1112335MCP
get_claim(id=1112335)