Have Multimodal Large Language Models (MLLMs) Really Learned to Tell the Time on Analog Clocks?

TL;DR

This paper investigates whether multimodal large language models like GPT-4.1 truly understand how to tell time on analog clocks or if they merely recognize patterns from training data, highlighting their limitations in generalization.

Contribution

The study evaluates GPT-4.1's ability to tell time on analog clocks, revealing its progress and limitations in understanding versus pattern recognition.

Findings

Models show some ability to read clock times.

Models struggle with generalizing to new clock images.

Fine-tuning may improve but does not fully solve the problem.

Abstract

Multimodal Large Language Models which can answer complex questions on an image struggle to tell the time on analog clocks. This is probably due to the lack of images with clocks at different times in their training set. In this work we explore this issue with one of the latest MLLMs: GPT-4.1 to understand why MLLMs fail to tell the time and whether fine-tuning can solve the problem. The results show how models are making progress in reading the time on analog clocks. But have they really learned to do it, or have they only learned patterns in their training datasets? In this work we put the models to the test with different clocks to illustrate the limitations of MLLMs to abstract and generalize.