# CountQA: How Well Do MLLMs Count in the Wild?

arXiv: https://arxiv.org/abs/2508.06585 · HTML: https://mlsanity.com/paper/2508.06585/

## Key facts

- Cited by 19 papers in the map (data through 14 Oct 2026).
- Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 21 of its 35 references that are in the map.
- First posted on arXiv 8 Aug 2025.

## Placed in

- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 21 of 21)

- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [You Only Look Once: Unified, Real-Time Object Detection](https://mlsanity.com/paper/1506.02640/) (2015)
- [Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection](https://mlsanity.com/paper/2303.05499/) (2023)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [Composition Loss for Counting, Density Map Estimation and Localization in Dense Crowds](https://mlsanity.com/paper/1808.01050/) (2018)
- [NWPU-Crowd: A Large-Scale Benchmark for Crowd Counting and Localization](https://mlsanity.com/paper/2001.03360/) (2020)
- [JHU-CROWD++: Large-Scale Crowd Counting Dataset and A Benchmark Method](https://mlsanity.com/paper/2004.03597/) (2020)
- [Vibe-Eval: A hard evaluation suite for measuring progress of multimodal language models](https://mlsanity.com/paper/2405.02287/) (2024)
- [Training-free Object Counting with Prompts](https://mlsanity.com/paper/2307.00038/) (2023)
- [LVLM-COUNT: Enhancing the Counting Ability of Large Vision-Language Models](https://mlsanity.com/paper/2412.00686/) (2024)
- [T2ICount: Enhancing Cross-modal Understanding for Zero-Shot Counting](https://mlsanity.com/paper/2502.20625/) (2025)
- [Math Blind: Failures in Diagram Understanding Undermine Reasoning in MLLMs](https://mlsanity.com/paper/2503.20745/) (2025)
- [GeoChain: Multimodal Chain-of-Thought for Geographic Reasoning](https://mlsanity.com/paper/2506.00785/) (2025)
- [MAEA: Multimodal Attribution for Embodied AI](https://mlsanity.com/paper/2307.13850/) (2023)
- [HueManity: Probing Fine-Grained Visual Perception in MLLMs](https://mlsanity.com/paper/2506.03194/) (2025)
- [Semantic Augmentation in Images using Language](https://mlsanity.com/paper/2404.02353/) (2024)
- [A Community-Centric Perspective for Characterizing and Detecting Anti-Asian Violence-Provoking Speech](https://mlsanity.com/paper/2407.15227/) (2024)
- [AI Guide Dog: Egocentric Path Prediction on Smartphone](https://mlsanity.com/paper/2501.07957/) (2025)
- [Evaluating and Advancing Multimodal Large Language Models in Perception Ability Lens](https://mlsanity.com/paper/2411.14725/) (2024)

## Built on by (most cited first, 19 of 19)

- [Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception](https://mlsanity.com/paper/2602.11858/) (2026)
- [STEP3-VL-10B Technical Report](https://mlsanity.com/paper/2601.09668/) (2026)
- [Vero: An Open RL Recipe for General Visual Reasoning](https://mlsanity.com/paper/2604.04917/) (2026)
- [Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving](https://mlsanity.com/paper/2609.00111/) (2026)
- [MaRVL-QA: A Benchmark for Mathematical Reasoning over Visual Landscapes](https://mlsanity.com/paper/2508.17180/) (2025)
- [The Artificial Intelligence Cognitive Examination: A Survey on the Evolution of Multimodal Evaluation from Recognition to Reasoning](https://mlsanity.com/paper/2510.04141/) (2025)
- [Explicit Reasoning Makes Better Judges: A Systematic Study on Accuracy, Efficiency, and Robustness](https://mlsanity.com/paper/2509.13332/) (2025)
- [Next-Gen CAPTCHAs: Leveraging the Cognitive Gap for Scalable and Diverse GUI-Agent Defense](https://mlsanity.com/paper/2602.09012/) (2026)
- [Scaling Native Multimodal Pre-Training From Scratch](https://mlsanity.com/paper/2607.22043/) (2026)
- [Thinking with Visual Grounding](https://mlsanity.com/paper/2606.16122/) (2026)
- [HoloCount: A Holistic Visual Counting Benchmark for MLLMs](https://mlsanity.com/paper/2607.06420/) (2026)
- [Trace: A Taxonomy-Guided Environment for Multidomain Visual Reasoning](https://mlsanity.com/paper/2607.19790/) (2026)
- [Hard Vision, Easy Vision: What GPT-6 Astra Reveals Across Computer Vision](https://mlsanity.com/paper/2609.35718/) (2026)
- [Architectural Sampling: Test-Time Scaling via Computational Diversity in Frozen Vision-Language Models](https://mlsanity.com/paper/2610.01687/) (2026)
- [Confidence Aware SSD Ensemble with Weighted Boxes Fusion for Weapon Detection](https://mlsanity.com/paper/2509.23697/) (2025)
- [ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and Generation](https://mlsanity.com/paper/2606.23835/) (2026)
- [Thinking with Anchors: Grounded and Efficient Document Reasoning](https://mlsanity.com/paper/2608.04424/) (2026)
- [Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes](https://mlsanity.com/paper/2610.02117/) (2026)
- [Why VLMs Miss Small Objects, and When Zooming In Is Safe](https://mlsanity.com/paper/2610.09313/) (2026)

## Abstract (arXiv)

Multimodal Large Language Models (MLLMs) demonstrate remarkable fluency in understanding visual scenes, yet they exhibit a critical lack in a fundamental cognitive skill: object counting. This blind spot severely limits their reliability in real-world applications. To date, this capability has been largely unevaluated in complex scenarios, as existing benchmarks either feature sparse object densities or are confined to specific visual domains, failing to test models under realistic conditions. Addressing this gap, we introduce CountQA, a challenging new benchmark designed to probe this deficiency. Comprising over 1,500 question-answer pairs, CountQA features real-world images with high object density, clutter, and occlusion. We investigate this weakness by evaluating 15 prominent MLLMs on the CountQA benchmark and reveal that the top-performing model achieves a mere 42.9% accuracy, with performance declining as object counts rise. By providing a dedicated benchmark to diagnose and rectify this core weakness, CountQA paves the way for a new generation of MLLMs that are not only descriptively fluent but also numerically grounded and spatially aware. We will open-source the dataset and code upon paper acceptance to foster further research.