Update dataset card metadata, paper link and usage

This PR improves the dataset card by:
1. Adding `task_categories: [other]` to the YAML metadata.
2. Fixing the broken paper link in the Markdown content.
3. Updating the sample usage command to match the instructions provided in the official GitHub repository.
4. Correcting a typo in the license reference (from `44.0` to `4.0`).

README.md

@@ -1,18 +1,20 @@
 ---
-license: cc-by-nc-sa-4.0
 language:
 - en
+license: cc-by-nc-sa-4.0
+task_categories:
+- other
 ---
 
 # OWM Benchmark
 
 [![Project Page](https://img.shields.io/badge/Project-Page-blue)](https://compvis.github.io/myriad)
-[![Paper](https://img.shields.io/badge/arXiv-paper-b31b1b)](https://arxiv.org/abs/2604.09527)
+[![Paper](https://img.shields.io/badge/arXiv-paper-b31b1b)](https://huggingface.co/papers/2604.09527)
 [![MYRIAD Weights](https://img.shields.io/badge/HuggingFace-Weights-orange)](https://huggingface.co/CompVis/myriad)
 
 ## Abstract
 
-The OWM benchmark was proposed in the paper [Envisioning the Future, One Step at a Time](_blank) and used to evaluate the [MYRIAD](https://huggingface.co/CompVis/myriad/) model.
+The OWM benchmark was proposed in the paper [Envisioning the Future, One Step at a Time](https://huggingface.co/papers/2604.09527) and used to evaluate the [MYRIAD](https://huggingface.co/CompVis/myriad/) model.
 
 OWM is a benchmark of 95 curated videos with motion annotations, with the distribution of motion constrained to enable the evaluation of probabilistic motion prediction methods.
 Videos are obtained from Pexels ([Pexels License](https://www.pexels.com/license/)). We manually annotate relevant objects and the type of motion observed. We use an off-the-shelf tracker to obtain motion trajectories and manually verify correctness.
@@ -24,21 +26,23 @@ Videos are obtained from Pexels ([Pexels License](https://www.pexels.com/license
 
 ![OWM samples](https://compvis.github.io/myriad/static/images/paper-svg/owm-qualitative.svg)
 
-*OWM samples are include complex real-world scenes with different motion types and complexities.*
+*OWM samples include complex real-world scenes with different motion types and complexities.*
 
 ## Usage
 
 We provide code to run the OWM evaluation in our [GitHub repository](https://github.com/CompVis/flow-poke-transformer).
 
-To run the evaluation, first download the data by running `hf download CompVis/owm-95 --repo-type dataset` then run the evaluation script via
+To run the evaluation, first download the data by running `hf download CompVis/owm-95 --repo-type dataset`.
+
+Then run the evaluation script via:
 ```shell
-python -m scripts.eval.myriad_eval.openset_prediction --checkpoint-path path/to/checkpoint  -data-path path/to/data
+python -m scripts.myriad_eval.openset_prediction --data-root path/to/data --ckpt-path path/to/checkpoint --dataset-name owm
 ```
 
 ## License
 
 - Videos are sourced from Pexels and thus licensed under the [Pexels License](https://www.pexels.com/license/)
-- Metadata and motion annotations are provided under the [CC-BY-NC-SA-44.0](https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en) license
+- Metadata and motion annotations are provided under the [CC-BY-NC-SA-4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en) license
 
 ## Citation