Understanding the Limitations of Mathematical Reasoning in Large Language Models

Posted by iforgotmysocks on Saturday October 12, 2024 @06:09PM
from the machinelearning.apple.com dept.

https://machinelearning.apple.com/research/gsm-symbolic

( 3 comments | 2 points | on HN )

Re:Understanding the Limitations of Mathematical Reasoning i... by belter on Sunday October 13, 2024 @05:37AM (#41826511)

"The question of whether a computer can think is no more interesting than the question of whether a submarine can swim." - Edsger Dijkstra

[ Parent | on HN ]

Re:Understanding the Limitations of Mathematical Reasoning i... by msoad on Monday October 14, 2024 @11:00AM (#41838178)

A calculator can "think" is "AI" and however you want to frame it. Reasoning is a very specific and defined concept. Computers can not reason (per this paper)

[ Parent | on HN ]

Re:Understanding the Limitations of Mathematical Reasoning i... by aktuel on Sunday October 13, 2024 @12:43PM (#41829449)

The paper confirms that even the biggest and most advanced models can still be misled by adding irrelevant information that can be easily filtered out by humans.

[ Parent | on HN ]