The fact that these models are more effective at providing scaffolding than at applying rigor seems a symptom of them being products made to please users. Forcing deeper reasoning tends to irritate the other person. It is also interesting the quote "While prompting encourages models to push for rigor, they use fewer strategies than humans do", the models are so tunned to please that they can't even be creative to provoke the user to think for themselves
Guilherme Alves
guilhermecxe
AI & ML interests
None yet
Recent Activity
commentedon an article about 1 month ago
TutorMoments: Do AI tutors know when to help and when to hold back? published a model 9 months ago
guilhermecxe/gemma-text-to-sql liked a model about 1 year ago
CEIA-UFG/Gemma-3-Gaia-PT-BR-4b-itOrganizations
None yet
commented on TutorMoments: Do AI tutors know when to help and when to hold back? about 1 month ago
published a
model 9 months ago
liked a
model about 1 year ago
commented on Open R1: How to use OlympicCoder locally for coding over 1 year ago