Verified70% confidenceFactExact time
Research 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.
4
Sources
70%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
AI真的聪明吗？下一代人工智能的发展方向与能力边界
hackernewshackernews7/3/2026
Related Claims
Unverified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83% 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.82% similarUnverifiedHallucination 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".79% similarUnverifiedLarge models have higher overall accuracy, but their errors are harder to detect because their output is more fluent and persuasive.78% similarVerifiedPure large language models have two major deficiencies: a knowledge cutoff problem and a hallucination problem.78% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/485454API
curl https://kongchang.com/api/v1/knowledge/claims/485454MCP
get_claim(id=485454)