Update README.md

README.md

@@ -20,7 +20,7 @@ To enhance efficiency, we replaced the causal self-attention layers with bidirec
 Finally, our implementation integrates FlashAttention V2 for faster inference. 
 
 
-- **Paper:** [Link](https://arxiv.org/abs/2503.03008)
+- **Paper:** [One Model to Train them All: Hierarchical Self-Distillation for Enhanced Early Layer Embeddings](https://arxiv.org/abs/2503.03008)
 - **Languages:** 600+ Programming languages