Update README.md

README.md

@@ -1,5 +1,17 @@
 ---
 license: apache-2.0
+datasets:
+- JoshuaW1997/FUTGA
+language:
+- en
+metrics:
+- bertscore
+- bleu
+- rouge
+base_model:
+- lmsys/vicuna-7b-v1.5
+tags:
+- music
 ---
 
 # FUTGA: Fine-grained Music Understanding through Temporally-enhanced Generative Augmentation
@@ -8,6 +20,7 @@ license: apache-2.0
 
 ## News
 
+- [10/02] We released [**FUTGA-Dataset**](https://huggingface.co/datasets/JoshuaW1997/FUTGA) including MusicCaps, SongDescriber, HarmonixSet, and AudioSet (train and eval).
 - [07/28] We released [**FUTGA-7B**](https://huggingface.co/JoshuaW1997/FUTGA) and **training/inference code** based on [SALMONN-7B](https://huggingface.co/tsinghua-ee/SALMONN) backbone!
 
 ## Overview
@@ -29,5 +42,14 @@ We build **FUTGA-7B** based on SALMONN. Follow the instructions from [SALMONN](h
 
 
 ## Datasets
-We generate dense captions for full-length songs in [MusicCaps](https://huggingface.co/JoshuaW1997/FUTGA/tree/main/Data-MusicCaps) and [SongDescriber](https://huggingface.co/JoshuaW1997/FUTGA/tree/main/Data-SongDescriber), 
-where ```raw captions``` are directly generated from FUTGA-7B and ```seg_captions_features``` contain automatically segmented captions with structures and textual-audio features for each segment.
+Please visit our dataset repo [FUTGA-Dataset](https://huggingface.co/datasets/JoshuaW1997/FUTGA). We currently include MusicCaps, SongDescriber, HarmonixSet, and AudioSet (train and eval).
+
+## Citation
+If you use our models or datasets in your research, please cite it as follows:
+```bibtex
+@article{wu2024futga,
+  title={Futga: Towards Fine-grained Music Understanding through Temporally-enhanced Generative Augmentation},
+  author={Wu, Junda and Novack, Zachary and Namburi, Amit and Dai, Jiaheng and Dong, Hao-Wen and Xie, Zhouhang and Chen, Carol and McAuley, Julian},
+  journal={arXiv preprint arXiv:2407.20445},
+  year={2024}
+}