Improve dataset card links and documentation

README.md

@@ -55,7 +55,7 @@ SA-BENCH is the benchmark dataset released with **“Beyond Pixels: Benchmarking
 
 **Accepted to CVPRW 2026.**
 
-Code: https://github.com/AlibabaResearch/SA-IQA
+[GitHub](https://github.com/AlibabaResearch/SA-IQA) | [CVF Open Access](https://openaccess.thecvf.com/content/CVPR2026W/NTIRE/html/Gao_Beyond_Pixels_Benchmarking_and_Reward-Based_Assessing_Framework_for_Visual_Spatial_CVPRW_2026_paper.html) | [arXiv](https://arxiv.org/abs/2512.05098) | [Model](https://huggingface.co/gaoyuan-ai/SA-IQA-model)
 
 It evaluates the spatial aesthetics of interior images along four dimensions:
 
@@ -64,7 +64,9 @@ It evaluates the spatial aesthetics of interior images along four dimensions:
 - **layout**
 - **lighting**
 
-The dataset contains **17,768 annotated examples** across four spatial-aesthetic dimensions, with image assets and human annotations for training and evaluation.
+SA-BENCH contains **17,768 annotated examples** across four spatial-aesthetic dimensions, with image assets and human annotations for training and evaluation.
+
+> **Dataset Viewer Note:** The Hugging Face Dataset Viewer may report a larger aggregate row count because `train`, `test`, and `full` CSV files are exposed as separate splits. The benchmark itself contains **17,768 examples**.
 
 ## Dataset Details