Update dataset card for VisNumBench: Correct content, add links, sample usage, and refine tags

README.md

@@ -1,5 +1,12 @@
 ---
 license: mit
+task_categories:
+- image-text-to-text
+tags:
+- multimodal
+- number-sense
+- visual-reasoning
+- benchmark
 configs:
 - config_name: default
   data_files:
@@ -29,29 +36,61 @@ dataset_info:
     num_examples: 1913
   download_size: 230897223
   dataset_size: 82349062.411
-task_categories:
-- image-text-to-text
-tags:
-- geometry
-- mathematical-reasoning
-- multimodal
 ---
 
-This dataset is designed for research in **Deep Learning for Geometry Problem Solving (DL4GPS)** and accompanies the survey paper [A Survey of Deep Learning for Geometry Problem Solving](https://huggingface.co/papers/2507.11936). It aims to provide a structured resource for evaluating and training AI models, particularly multimodal large language models (MLLMs), on mathematical reasoning tasks involving geometric contexts.
+# VisNumBench: Evaluating Number Sense of Multimodal Large Language Models
+
+This repository contains the official evaluation code and data for **VisNumBench: Evaluating Number Sense of Multimodal Large Language Models**.
+
+**Paper:** [VisNumBench: Evaluating Number Sense of Multimodal Large Language Models](https://huggingface.co/papers/2503.14939)
+**Project Homepage:** https://wwwtttjjj.github.io/VisNumBench/
+**Code:** https://github.com/wwwtttjjj/VisNumBench
+
+## Introduction
 
-The dataset provides a collection of geometry problems, each consisting of a textual question and a corresponding image.
+Can Multimodal Large Language Models (MLLMs) develop an intuitive number sense similar to humans? Targeting this problem, we introduce Visual Number Benchmark (<b>VisNumBench</b>) to evaluate the number sense abilities of MLLMs across a wide range of visual numerical tasks. <b>VisNumBench</b> consists of about 1,900 multiple-choice question-answer pairs derived from both synthetic and real-world visual data, covering seven visual numerical attributes and four types of visual numerical estimation tasks. Our experiments on <b>VisNumBench</b> led to the following key findings: (i) The 17 MLLMs we tested—including open-source models such as Qwen2.5-VL and InternVL2.5, as well as proprietary models like GPT-4o and Gemini 2.0 Flash—perform significantly below human levels in number sense-related tasks. (ii) Multimodal mathematical models and multimodal chain-of-thought (CoT) models did not exhibit significant improvements in number sense abilities. (iii) Stronger MLLMs with larger parameter sizes and broader general abilities demonstrate modest gains in number sense abilities. We believe <b>VisNumBench</b> will serve as a valuable resource for the research community, encouraging further advancements in enhancing LVLMs' number sense abilities.
 
-For a continuously updated reading list of papers on Deep Learning for Geometry Problem Solving, refer to the [official GitHub repository](https://github.com/majianz/gps-survey).
+## Dataset Creation
+
+VisNumBench aims to advance the development of multimodal large language models in visual numerical understanding by evaluating their number sense capabilities. This benchmark is dedicated to bridging the gap between abstract mathematical problem-solving and real-world applications in current multimodal models.
 
 ## Data Structure
 
 Each problem instance in the dataset includes the following fields:
 
--   `class`: The category of the geometry problem.
+-   `class`: The category of the visual number problem.
 -   `id`: A unique identifier for each problem.
--   `question`: The textual description of the geometry problem.
--   `option`: Multiple-choice options for the answer, if applicable.
--   `answer`: The correct answer to the geometry problem.
--   `task_class`: A classification of the task involved.
--   `Attributes`: Additional attributes or metadata about the problem.
--   `image`: The image of the geometric diagram associated with the problem.
+-   `question`: The textual question related to the visual numerical task.
+-   `option`: Multiple-choice options for the answer.
+-   `answer`: The correct answer to the problem.
+-   `task_class`: A classification of the task involved, such as `Range Estimation`, `Value Comparison`, `Value Estimation`, or `Multiplicative Estimation`.
+-   `Attributes`: Visual numerical attributes covered, including `Angle`, `Length`, `Scale`, `Depth`, `Quantity`, `Area`, and `Volume`.
+-   `image`: The visual data (image) associated with the problem.
+
+## Load Dataset
+
+You can load the dataset using the Hugging Face `datasets` library:
+
+```python
+from datasets import load_dataset
+
+# Login using e.g. `huggingface-cli login` to access this dataset
+ds = load_dataset("GML-FMGroup/VisNumBench")
+```
+
+## Evaluation
+
+Please refer to the [evaluation folder](https://github.com/wwwtttjjj/VisNumBench/tree/main/eval) in the GitHub repository for more details on evaluation.
+
+## Citation
+
+If you use VisNumBench in your research, please cite the following paper:
+
+```bibtex
+@inproceedings{weng2025visnumbench,
+  title={VisNumBench: Evaluating Number Sense of Multimodal Large Language Models},
+  author={Tengjin Weng and Wenhao Jiang and Jingyi Wang and Zhong Ming},
+  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
+  year={2025}
+}
+```