Add model card, link to Github and paper

This PR improves the model card with:
- the appropriate `pipeline_tag`
- a link to the paper

README.md

@@ -1,5 +1,11 @@
 ---
 license: apache-2.0
+pipeline_tag: video-audio-to-text
+library_name: transformers
 ---
-Pretrain and finetune weights for CVPR'25 Paper "Crab: A Unified Audio-Visual Scene Understanding Model
-with Explicit Cooperation"
+
+This repository contains the pretrain and finetune weights for the model introduced in the paper "Crab: A Unified Audio-Visual Scene Understanding Model with Explicit Cooperation".
+
+[Paper](https://huggingface.co/papers/2503.13068)
+
+Github: https://github.com/GeWu-Lab/Crab