Add task categories and link to paper

README.md

@@ -1,23 +1,25 @@
 ---
-license: apache-2.0
 language:
 - en
+license: apache-2.0
+size_categories:
+- 1M<n<10M
+task_categories:
+- graph-ml
 tags:
 - graph
 - text-attributed-graph
 - node-classification
 - llm
-size_categories:
-- 1M<n<10M
 ---
 
 # GraspLLM Datasets
 
-Pre-processed text-attributed graphs used in
-[**GraspLLM**](https://github.com/Heinz217/GraspLLM):
-*Towards Zero-Shot Generalization on Text-Attributed Graphs with LLMs*.
+Pre-processed text-attributed graphs used in the paper [**GraspLLM: Towards Zero-Shot Generalization on Text-Attributed Graphs with LLMs**](https://huggingface.co/papers/2606.11898).
+
+**Official Code Repository**: [https://github.com/Heinz217/GraspLLM](https://github.com/Heinz217/GraspLLM)
 
-This repository hosts the **raw graph** (`processed_data.pt`) for each dataset.
+This repository hosts the **raw graph** (`processed_data.pt`) for each dataset. GraspLLM is a framework that combines graph structural comprehension with the semantic understanding prowess of LLMs, enabling zero-shot reasoning over Text-Attributed Graphs (TAGs).
 
 ## Datasets
 
@@ -42,6 +44,8 @@ Each `<dataset>/processed_data.pt` is a [PyG](https://pytorch-geometric.readthed
 
 ## Usage
 
+You can download the dataset bundle using the Hugging Face CLI:
+
 ```bash
 # Clone the GraspLLM code repo
 git clone https://github.com/Heinz217/GraspLLM.git
@@ -55,3 +59,24 @@ huggingface-cli download Heinz217/GraspLLM-Datasets \
 # Configure dataset root (or just keep the default ./dataset)
 export GRASPLLM_DATASET_ROOT=$PWD/dataset
 ```
+
+The expected layout for each dataset in the pipeline is:
+
+```
+$GRASPLLM_DATASET_ROOT/<name>/
+├── processed_data.pt        # provided here
+├── qwen3_emb_x.pt           # produced by Step 0 
+├── ocs_train.jsonl          # produced by Stage 2
+└── ocs_test.jsonl           # produced by Stage 2
+```
+
+## Citation
+
+```bibtex
+@article{graspllm2026,
+  title={GraspLLM: Towards Zero-Shot Generalization on Text-Attributed Graphs with LLMs},
+  author={Zhou, Heinz and others},
+  journal={arXiv preprint arXiv:2606.11898},
+  year={2026}
+}
+```