Update README.md

README.md

@@ -18,13 +18,12 @@ size_categories:
 - 100K<n<1M
 ---
 
-<div align="center">
+
 
 # 🧬 ScDiVa Benchmark Datasets
 
 [**📄 arXiv Paper**](https://arxiv.org/abs/2602.03477) | [**💻 Model Repository**](https://huggingface.co/warming666/ScDiVa) | [**🌐 Project Page**](https://github.com/warming666/ScDiVa)
 
-</div>
 
 ## 📖 Dataset Description
 
@@ -85,13 +84,15 @@ The datasets are categorized by their primary evaluation task in the ScDiVa pape
 | `adamson_processed.h5ad` | Single-gene CRISPRi perturbations (Adamson et al., 2016). |
 | `norman.h5ad` | Combinatorial (double-gene) CRISPRa perturbations (Norman et al., 2019). |
 
-### 5. Reconstruction & GRN
+### 5. Reconstruction & GRN Inference
+*Datasets used to evaluate the model's ability to reconstruct gene expression (Section 4.2) and infer regulatory networks (Section 4.6).*
+
 | Filename | Description |
 | :--- | :--- |
-| `zheng68k_processed.h5ad` | **Zheng68k**: Large-scale PBMC dataset for reconstruction and GRN. |
-| `immune_processed.h5ad` | **Immune Atlas**: Used for high-complexity reconstruction benchmarks. |
-| `hpancreas_processed.h5ad` | **hPancreas**: Used for evaluating reconstruction on distinct cell types. |
-| `pbmc12k_processed.h5ad` | **PBMC12k**: Used for evaluating reconstruction fidelity. |
+| `zheng68k_processed.h5ad` | Mouse hematopoietic stem and progenitor cell scRNA-seq dataset (68k cells). |
+| `hpancreas_processed.h5ad` | Human Pancreas dataset. |
+| `pbmc12k_processed.h5ad` | PBMC dataset with 2 batches. |
+| `immune_processed.h5ad` | Human Immune Atlas(Primary dataset for GRN inference). |
 
 ---