Improve model card: Add pipeline tag and paper link

This PR enhances the model card for VFRTok by:
- Adding `pipeline_tag: other` to the metadata, which helps users discover this specialized video tokenizer on the Hugging Face Hub.
- Including an explicit link to the paper, [VFRTok: Variable Frame Rates Video Tokenizer with Duration-Proportional Information Assumption](https://huggingface.co/papers/2505.12053), in the content for easier access and proper attribution.

These updates improve the model's discoverability and documentation without altering existing functional elements or adding library compatibility where no code evidence was found.

README.md

@@ -1,6 +1,8 @@
 ---
 license: mit
+pipeline_tag: other
 ---
+
 # [NeurIPS'25] VFRTok: Variable Frame Rates Video Tokenizer with Duration-Proportional Information Assumption
 
 <div align="center">
@@ -15,6 +17,8 @@ license: mit
 <img src="assets/teaser.png" alt="Teaser" width="512">
 </div>
 
+This model was presented in the paper [VFRTok: Variable Frame Rates Video Tokenizer with Duration-Proportional Information Assumption](https://huggingface.co/papers/2505.12053).
+
 ## Get Started
 ### Download Pretrain Models
 The weights `vfrtok-l.bin` and `vfrtok-s.bin` can be downloaded [here](https://huggingface.co/KwaiVGI/VFRTok).