Add dataset card for Vision-Zero CLEVR dataset

README.md

@@ -0,0 +1,80 @@
+---
+task_categories:
+- image-text-to-text
+- question-answering
+language:
+- en
+tags:
+- vlm
+- self-improvement
+- gamified-self-play
+- reinforcement-learning
+- clevr
+---
+
+# Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play (CLEVR-based Dataset)
+
+This repository contains the CLEVR-based dataset used in the paper [Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play](https://huggingface.co/papers/2509.25541).
+
+The Vision-Zero framework proposes VLM self-improvement through competitive visual games. This specific dataset comprises CLEVR-based synthetic scenes, serving as one of the key data types used to demonstrate the framework's versatility and enhance the reasoning abilities of Vision-Language Models.
+
+**Code Repository:** [https://github.com/wangqinsi1/Vision-Zero](https://github.com/wangqinsi1/Vision-Zero)
+
+## Abstract
+Although reinforcement learning (RL) can effectively enhance the reasoning capabilities of vision-language models (VLMs), current methods remain heavily dependent on labor-intensive datasets that require extensive manual construction and verification, leading to extremely high training costs and consequently constraining the practical deployment of VLMs. To address this challenge, we propose Vision-Zero, a domain-agnostic framework enabling VLM self-improvement through competitive visual games generated from arbitrary image pairs. Specifically, Vision-Zero encompasses three main attributes: (1) Strategic Self-Play Framework: Vision-Zero trains VLMs in "Who Is the Spy"-style games, where the models engage in strategic reasoning and actions across multiple roles. Through interactive gameplay, models autonomously generate their training data without human annotation. (2) Gameplay from Arbitrary Images: Unlike existing gamified frameworks, Vision-Zero can generate games from arbitrary images, thereby enhancing the model's reasoning ability across diverse domains and showing strong generalization to different tasks. We demonstrate this versatility using three distinct types of image datasets: CLEVR-based synthetic scenes, charts, and real-world images. (3) Sustainable Performance Gain: We introduce Iterative Self-Play Policy Optimization (Iterative-SPO), a novel training algorithm that alternates between Self-Play and reinforcement learning with verifiable rewards (RLVR), mitigating the performance plateau often seen in self-play-only training and achieving sustained long-term improvements. Despite using label-free data, Vision-Zero achieves state-of-the-art performance on reasoning, chart question answering, and vision-centric understanding tasks, surpassing other annotation-based methods. Models and code has been released at this https URL.
+
+## Sample Usage (Training)
+
+To use this CLEVR-based dataset within the Vision-Zero framework for training, follow these steps to set up your environment and prepare the dataset configuration.
+
+### 🛠️ Setup
+
+First, create a `conda` environment and install the necessary dependencies:
+
+```bash
+conda create -n vision-zero python=3.10
+conda activate vision-zero
+bash setup.sh
+```
+
+### 💪🏻 Training
+
+#### Step 1: Prepare Dataset and Model
+
+You can use this Clevr-based dataset. Update the model, dataset paths, and output address in `run_scripts/run_grpo_vision_zero.sh`:
+
+```python
+IMAGES_DIR = $IMAGES_DIR  # Path to your images (e.g., from this dataset)
+SCENES_DIR = $SCENES_DIR  # Path to your scenes (e.g., from this dataset)
+MODEL = $MODEL            # Your chosen VLM model (e.g., Qwen2.5-VL-7B)
+OUTPUT_BASE_DIR = $OUTPUT_DIR
+RUN_NAME= [run_name]
+```
+
+#### Step 2: Training Model
+
+Use the following command to train the model. You can modify all hyperparameters in `run_scripts/run_grpo_vision_zero.sh`.
+
+```bash
+bash run_scripts/run_grpo_vision_zero.sh
+```
+
+#### Step 3: Evaluation
+
+After training and saving the checkpoint, you can use [VLMEvalKit](https://github.com/open-compass/VLMEvalKit) to test the model's performance on out-of-distribution (OOD) tasks, as described in the original project.
+
+## Citation
+
+If you find Vision-Zero useful or relevant to your project and research, please kindly cite our paper:
+
+```bibtex
+@misc{wang2025visionzeroscalablevlmselfimprovement,
+      title={Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play}, 
+      author={Qinsi Wang and Bo Liu and Tianyi Zhou and Jing Shi and Yueqian Lin and Yiran Chen and Hai Helen Li and Kun Wan and Wentian Zhao},
+      year={2025},
+      eprint={2509.25541},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2509.25541}, 
+}
+```