Fix YAML metadata frontmatter, add paper and GitHub links

README.md

@@ -1,46 +1,46 @@
----
-license: apache-2.0
-task_categories:
-- image-classification
-tags:
-- AIGC-detection
-- image-forensics
-- MLLM-Generated
-pretty_name: MLLM-Generated Image Dataset
----
-
-# MLLM-Generated Image Dataset
-
-This dataset contains real and AI-generated image samples organized for binary MLLM-generated image detection.
-
-## Dataset Summary
-
-We construct an MLLM-generated image detection benchmark from GPT Image2 and Nano Banana2. This benchmark covers texture-dominated, structure-dominated, and hybrid-dominated. It is designed to evaluate detector performance under the new challenges introduced by large-scale image generation models.
-
-## Data Fields
-
-- `file_name`: relative path to the image file.
-- `label`: integer class label, where `0` is real and `1` is fake.
-- `label_name`: string label name.
-- `domain`: image domain, such as `hybrid`, `structure`, or `texture`.
-- `generator`: source generator name, or `real` for authentic images.
-- `source_folder`: original top-level folder name.
-- `original_file_name`: original image file name.
-
-## Usage
-
-```python
-from datasets import load_dataset
-
-dataset = load_dataset("zr-zhang/MLLM-Generated-Image-Detection-Dataset")
-```
-
-## Licensing and Restrictions
-
-This dataset is made available solely for research on AI-generated content detection, image authenticity analysis, digital forensics, generative model evaluation, and related academic purposes. It must not be used to develop or optimize technologies for fraud, forgery, identity impersonation, fabrication of false evidence, or circumvention of content detection systems.
-
-
-## Citation
-
-The associated paper is currently under review. Citation information will be updated once the paper becomes publicly available.
-
+---
+license: apache-2.0
+task_categories:
+- image-classification
+tags:
+- AIGC-detection
+- image-forensics
+- MLLM-Generated
+pretty_name: MLLM-Generated Image Dataset
+---
+
+# MLLM-Generated Image Dataset
+
+This dataset contains real and AI-generated image samples organized for binary MLLM-generated image detection.
+
+[Paper](https://huggingface.co/papers/2608.01258) | [Code](https://github.com/xbrainnet/SAP-DSP)
+
+## Dataset Summary
+
+We construct an MLLM-generated image detection benchmark from GPT Image2 and Nano Banana2. This benchmark covers texture-dominated, structure-dominated, and hybrid-dominated. It is designed to evaluate detector performance under the new challenges introduced by large-scale image generation models.
+
+## Data Fields
+
+- `file_name`: relative path to the image file.
+- `label`: integer class label, where `0` is real and `1` is fake.
+- `label_name`: string label name.
+- `domain`: image domain, such as `hybrid`, `structure`, or `texture`.
+- `generator`: source generator name, or `real` for authentic images.
+- `source_folder`: original top-level folder name.
+- `original_file_name`: original image file name.
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("zr-zhang/MLLM-Generated-Image-Detection-Dataset")
+```
+
+## Licensing and Restrictions
+
+This dataset is made available solely for research on AI-generated content detection, image authenticity analysis, digital forensics, generative model evaluation, and related academic purposes. It must not be used to develop or optimize technologies for fraud, forgery, identity impersonation, fabrication of false evidence, or circumvention of content detection systems.
+
+## Citation
+
+Please cite the paper [A Benchmark Dataset for MLLM-Generated Image Detection: GPT Image2 & Nano Banana2](https://huggingface.co/papers/2608.01258) when referencing or using this dataset.