# Can CLIP Count Stars? An Empirical Study on Quantity Bias in CLIP

arXiv: https://arxiv.org/abs/2409.15035 · HTML: https://mlsanity.com/paper/2409.15035/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 13 of its 22 references that are in the map.
- First posted on arXiv 23 Sep 2024.

## Builds on (most cited first, 13 of 13)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [An Investigation of Why Overparameterization Exacerbates Spurious Correlations](https://mlsanity.com/paper/2005.04345/) (2020)
- [OBJECT 3DIT: Language-guided 3D-aware Image Editing](https://mlsanity.com/paper/2307.11073/) (2023)
- [TexSliders: Diffusion-Based Texture Editing in CLIP Space](https://mlsanity.com/paper/2405.00672/) (2024)
- [Assessing Shape Bias Property of Convolutional Neural Networks](https://mlsanity.com/paper/1803.07739/) (2018)
- [Who's in and who's out? A case study of multimodal CLIP-filtering in DataComp](https://mlsanity.com/paper/2405.08209/) (2024)
- [The CLIP Model is Secretly an Image-to-Prompt Converter](https://mlsanity.com/paper/2305.12716/) (2023)
- [Isotropic3D: Image-to-3D Generation Based on a Single CLIP Embedding](https://mlsanity.com/paper/2403.10395/) (2024)
- [Bag of Tricks to Boost Adversarial Transferability](https://mlsanity.com/paper/2401.08734/) (2024)
- [Reducing Texture Bias of Deep Neural Networks via Edge Enhancing Diffusion](https://mlsanity.com/paper/2402.09530/) (2024)
- [Discover and Mitigate Multiple Biased Subgroups in Image Classifiers](https://mlsanity.com/paper/2403.12777/) (2024)

## Built on by (most cited first, 9 of 9)

- [Your Vision-Language Model Can't Even Count to 20: Exposing the Failures of VLMs in Compositional Counting](https://mlsanity.com/paper/2510.04401/) (2025)
- [On the Limitations of Vision-Language Models in Understanding Image Transforms](https://mlsanity.com/paper/2503.09837/) (2025)
- [Targeted Forgetting of Image Subgroups in CLIP Models](https://mlsanity.com/paper/2506.03117/) (2025)
- [Shape Bias and Robustness Evaluation via Cue Decomposition for Image Classification and Segmentation](https://mlsanity.com/paper/2503.12453/) (2025)
- [LLMs Can Compensate for Deficiencies in Visual Representations](https://mlsanity.com/paper/2506.05439/) (2025)
- [Will the Inclusion of Generated Data Amplify Bias Across Generations in Future Image Classification Models?](https://mlsanity.com/paper/2410.10160/) (2024)
- [Can Vision-Language Models See Squares? Text-Recognition Mediates Spatial Reasoning Across Three Model Families](https://mlsanity.com/paper/2602.15950/) (2026)
- [Why Instruction-Based Unlearning Fails in Diffusion Models?](https://mlsanity.com/paper/2604.01514/) (2026)
- [Is CLIP Cross-Eyed? Revealing and Mitigating Center Bias in the CLIP Family](https://mlsanity.com/paper/2604.05971/) (2026)

## Abstract (arXiv)

CLIP has demonstrated great versatility in adapting to various downstream tasks, such as image editing and generation, visual question answering, and video understanding. However, CLIP-based applications often suffer from misunderstandings regarding user intent, leading to discrepancies between the required number of objects and the actual outputs in image generation tasks. In this work, we empirically investigate the quantity bias in CLIP. By carefully designing different experimental settings and datasets, we comprehensively evaluate CLIP's understanding of quantity from text, image, and cross-modal perspectives. Our experimental results reveal a quantity bias in CLIP embeddings, impacting the reliability of downstream tasks.