Updates citation format and adds acknowledgment

Converts BibTeX citation to proper arXiv format with eprint ID and classification

Adds OpenSeek to acknowledgments for kernel development support

Includes additional author Liangdong Wang and Guang Liu in citation

Author: LoserCheems

README.md

@@ -235,18 +235,22 @@ This project is licensed under the BSD 3-Clause License. See [LICENSE](LICENSE)
 If you use Flash-DMA in your research, please cite:
 
 ```bibtex
-@misc{flash_dma_2025,
-  title={Trainable Dynamic Mask Sparse Attention},
-  author={Jingze Shi and Yifan Wu and Bingheng Wu and Yiran Peng and Yuyu Luo},
-  year={2025},
-  url={https://github.com/SmallDoges/flash-dmattn}
+@misc{shi2025trainabledynamicmasksparse,
+      title={Trainable Dynamic Mask Sparse Attention}, 
+      author={Jingze Shi and Yifan Wu and Bingheng Wu and Yiran Peng and Liangdong Wang and Guang Liu and Yuyu Luo},
+      year={2025},
+      eprint={2508.02124},
+      archivePrefix={arXiv},
+      primaryClass={cs.AI},
+      url={https://arxiv.org/abs/2508.02124}, 
 }
 ```
 
 ## Acknowledgments
 
 This project builds upon and integrates several excellent works:
 
+- **[OpenSeek](https://github.com/FlagAI-Open/OpenSeek)** - Kernel development support
 - **[Flash-Attention](https://github.com/Dao-AILab/flash-attention)** - Memory-efficient attention computation
 - **[NVIDIA CUTLASS](https://github.com/NVIDIA/cutlass)** - High-performance matrix operations library