Beyond Memorization: A Multi-Modal Ordinal Regression Benchmark to Expose Popularity Bias in Vision-Language Models
This work addresses a critical flaw in vision-language models' reasoning capabilities, which is important for researchers and developers in AI, though it is incremental in exposing a known bias.
The paper exposes a significant popularity bias in vision-language models, showing up to 34% higher accuracy on famous buildings compared to ordinary ones, and introduces the YearGuessr dataset with 55,546 images to benchmark this issue.
We expose a significant popularity bias in state-of-the-art vision-language models (VLMs), which achieve up to 34% higher accuracy on famous buildings compared to ordinary ones, indicating a reliance on memorization over generalizable understanding. To systematically investigate this, we introduce the largest open benchmark for this task: the YearGuessr dataset, a collection of 55,546 building images with multi-modal attributes from 157 countries, annotated with continuous ordinal labels of their construction year (1001-2024), GPS data, and page-view counts as a proxy for popularity. Using this dataset, we frame the construction year prediction task as ordinal regression and introduce popularity-aware interval accuracy metrics to quantify this bias. Our resulting benchmark of 30+ models, including our YearCLIP model, confirms that VLMs excel on popular, memorized items but struggle significantly with unrecognized subjects, exposing a critical flaw in their reasoning capabilities. Project page: https://sytwu.github.io/BeyondMemo/