Update README.md

README.md

@@ -32,7 +32,7 @@ KoVRE is a 2B-parameter single-vector embedding model designed for **Korean visu
 KoVRE is initialized from [`Qwen/Qwen3-VL-Embedding-2B`](https://huggingface.co/Qwen/Qwen3-VL-Embedding-2B) and trained in two stages. The first stage performs contrastive learning on 708,729 Korean and English query-page pairs with positive-aware hard-negative mining. The second stage distills relevance scores from `Qwen3-VL-Reranker-8B` using Korean data. Matryoshka Representation Learning enables flexible embedding dimensions from 128 to 2,048.
 
 - **Code:** https://github.com/whybe-choi/kovre
-- **Paper:** WIP
+- **Paper:** https://arxiv.org/abs/2608.01389
 
 ## Highlights
 
@@ -327,9 +327,10 @@ KoVRE is released under the Apache 2.0 License.
 If you find KoVRE useful, please cite:
 
 ```bibtex
-@misc{choi2026kovre,
-  title        = {KoVRE: Training an Efficient Embedding Model for Korean Visual Document Retrieval},
-  author       = {Yongbin Choi and Gyuho Shim and Youngjoon Jang},
-  year         = {2026}
+@article{choi2026kovre,
+  title={KoVRE: Training an Efficient Embedding Model for Korean Visual Document Retrieval},
+  author={Choi, Yongbin and Shim, Gyuho and Jang, Youngjoon},
+  journal={arXiv preprint arXiv:2608.01389},
+  year={2026}
 }
 ```