Add paper link, GitHub link, and task category to dataset card

README.md

@@ -1,19 +1,23 @@
 ---
-license: mit
 language:
 - en
 - zh
+license: mit
+task_categories:
+- video-text-to-text
 tags:
 - streaming
-- spatial Intelligence
+- spatial-intelligence
+- egocentric-video
 ---
 
-
 # UCS-Bench
 
+[**Paper**](https://huggingface.co/papers/2606.15200) | [**Code**](https://github.com/cocowy1/UCS-Bench)
+
 UCS-Bench is a benchmark for evaluating **user-centric continual spatial intelligence** in streaming egocentric videos.
 
-The dataset contains **170+ hours of egocentric visual observations** and **7K+ timestamped questions**. It is designed to test whether models can understand dynamic spatial environments from a user's point of view, maintain long-term spatial memory, and answer questions grounded in the user's real-time movement and location.
+The dataset contains **170+ hours of egocentric visual observations** and **8.1K+ timestamped questions**. It is designed to test whether models can understand dynamic spatial environments from a user's point of view, maintain long-term spatial memory, and answer questions grounded in the user's real-time movement and location.
 
 Unlike standard video QA benchmarks, UCS-Bench focuses on long-horizon spatial reasoning in continuous egocentric streams, including:
 
@@ -23,6 +27,18 @@ Unlike standard video QA benchmarks, UCS-Bench focuses on long-horizon spatial r
 * reasoning about dynamic scenes from the user's perspective;
 * aligning spatial memory with the user's changing location.
 
+## Dataset Download
+
+You can download the dataset using the `huggingface_hub` CLI:
+
+```bash
+pip install -U huggingface_hub
+
+huggingface-cli download cocowy1/UCS-Bench \
+  --repo-type dataset \
+  --local-dir data/UCS-Bench
+```
+
 ## Dataset Format
 
 Each video is paired with metadata in `metadata.jsonl`.
@@ -39,10 +55,6 @@ Fields:
 * `text`: question, caption, or annotation text
 * `label`: answer, category, or target label
 
-## Intended Use
-
-UCS-Bench is intended for research on egocentric AI assistants, streaming video understanding, spatial memory, and long-term multimodal reasoning.
-
 ## Source Datasets
 
 UCS-Bench is built upon and curated from several public egocentric and spatial understanding datasets. We gratefully acknowledge the original datasets and their contributors:
@@ -58,4 +70,13 @@ Please refer to the original dataset pages for their licenses, terms of use, and
 
 ## Citation
 
-Citation information will be added upon release.
+If you find UCS-Bench or DirectMe useful, please cite our work:
+
+```bibtex
+@misc{ucsbench2026,
+  title = {Keep It in Mind: User-Centric Continual Spatial Intelligence Reasoning in Egocentric Video Streams},
+  year = {2026},
+  note = {ICML 2026},
+  url = {https://icml.cc/virtual/2026/poster/63682}
+}
+```