See every side of every news story
Published loading...Updated

Cheat Questions to "Go Catch" Llms: Even the Best Fail with Such Trivial Questions

Summary by microsiervos.com
Next, a list of some of the typical questions with which you can quickly test any new version of an LLM and make it fail miserably, questioning whether it has really improved, is as powerful as they say or what noses happens in those "black boxes." Eye that even ChatGPT-5 seems to be failing in one of them, however infantile they seem, specifically in the numerical comparison and in the "blueberry test", although it depends on who tells you, bec…
DisclaimerThis story is only covered by news sources that have yet to be evaluated by the independent media monitoring agencies we use to assess the quality and reliability of news outlets on our platform. Learn more here.

1 Articles

Next, a list of some of the typical questions with which you can quickly test any new version of an LLM and make it fail miserably, questioning whether it has really improved, is as powerful as they say or what noses happens in those "black boxes." Eye that even ChatGPT-5 seems to be failing in one of them, however infantile they seem, specifically in the numerical comparison and in the "blueberry test", although it depends on who tells you, bec…

Think freely.Subscribe and get full access to Ground NewsSubscriptions start at $9.99/yearSubscribe

Bias Distribution

  • There is no tracked Bias information for the sources covering this story.

Factuality 

To view factuality data please Upgrade to Premium

Ownership

To view ownership data please Upgrade to Vantage

microsiervos.com broke the news in on Saturday, August 9, 2025.
Sources are mostly out of (0)