Improve dataset card: add paper link, code link, sample usage, and video-text-to-text task category

This PR improves the dataset card by:
- Adding a link to the paper on the Hugging Face hub.
- Adding a link to the official GitHub repository.
- Adding the `video-text-to-text` task category to the metadata, while keeping the existing task categories and tags.
- Adding a sample usage section based on the quick-start commands from the GitHub README.

README.md

@@ -1,11 +1,12 @@
 ---
+language:
+- en
 license: cc-by-nc-4.0
-pretty_name: RoboJudge-Bench
 task_categories:
 - video-classification
 - visual-question-answering
-language:
-- en
+- video-text-to-text
+pretty_name: RoboJudge-Bench
 tags:
 - robojudge
 - embodied-ai
@@ -15,10 +16,11 @@ tags:
 
 # RoboJudge-Bench
 
-RoboJudge-Bench contains 800 human-annotated embodied videos from 8 source
-corpora and 10 video generators. Every record includes the task instruction,
-locally hosted video and initial-frame URLs, PA/IA ratings, and six diagnostic
-sub-scores.
+**Paper:** [RoboJudge: Multimodal Language Models as Judges for Embodied Video Generation](https://huggingface.co/papers/2610.27782)  
+**Code:** [GitHub - Sphere-AI-Lab/RoboJudge](https://github.com/Sphere-AI-Lab/RoboJudge)  
+**Project page:** https://siyuanmacs.github.io/RoboJudge/
+
+RoboJudge-Bench contains 800 human-annotated embodied videos from 8 source corpora and 10 video generators. Every record includes the task instruction, locally hosted video and initial-frame URLs, PA/IA ratings, and six diagnostic sub-scores.
 
 ## Files
 
@@ -29,12 +31,24 @@ sub-scores.
 
 ## Scores
 
-- **Physical Adherence (PA):** agent consistency, scene consistency, and
-  interaction realism.
-- **Instruction Alignment (IA):** agent match, object correctness, and goal
-  completion.
+- **Physical Adherence (PA):** agent consistency, scene consistency, and interaction realism.
+- **Instruction Alignment (IA):** agent match, object correctness, and goal completion.
+
+Main scores are integers from 1 to 5. Diagnostic sub-scores are integers from 0 to 2.
+
+## Sample Usage
+
+```bash
+python -m pip install -r requirements.txt
+huggingface-cli download HuggingFriends/RoboJudge-Bench \
+  --repo-type dataset --local-dir data_root
 
-Main scores are integers from 1 to 5. Diagnostic sub-scores are integers from
-0 to 2.
+CKPT=HuggingFriends/RoboJudge-9B \
+DATA_ROOT="$PWD/data_root" \
+LIMIT=4 \
+bash code/run_inference.sh
 
-Project page: https://siyuanmacs.github.io/RoboJudge/
+python code/evaluate.py \
+  outputs/predictions.jsonl \
+  data/benchmark/robojudge_bench.jsonl
+```