News for hackers. Stuff that matters.

A comment on GPT-5. Parent comment

Re:GPT-5 by croemer on Thursday August 07, 2025 @01:34PM (#44827598)
Yes this is quite shocking. They could have just had o3 fact check the slides and it would have noticed...
[ Parent | on HN ]
Re:GPT-5 by throwaway0123_5 on Thursday August 07, 2025 @01:50PM (#44827864)
I thought so too, but I gave it a screenshot with the prompt:

> good plot for my presentation?

and it didn't pick up on the issue. Part of its response was:

> Clear metric: Y-axis (“Accuracy (%), pass @1”) and numeric labels make the performance gaps explicit.

I think visual reasoning is still pretty far from text-only reasoning.

[ Parent | on HN ]
Re:GPT-5 by abirch on Thursday August 07, 2025 @01:40PM (#44827687)
o3 did fact check the slides and it fixed its lower score.
[ Parent | on HN ]