Add benchmark tables and link TatBLiMP to its dataset

README.md

@@ -92,7 +92,7 @@ Ilshat Saetov, Dmitry Gaynullin.
 The training corpus is withheld on stewardship grounds for a minoritized language's literary heritage; a full datasheet with per-document provenance is available on reasonable request. The tokenizer and the model weights are released.
 
 ## License
-The model weights and the tokenizer are released under **CC BY-NC-SA 4.0**. The accompanying TatBLiMP benchmark is released under **CC BY-NC 4.0**. The training corpus is not released.
+The model weights and the tokenizer are released under **CC BY-NC-SA 4.0**. The accompanying [TatBLiMP](https://huggingface.co/datasets/yasalma/TatBLiMP) benchmark is released under **CC BY-NC 4.0**. The training corpus is not released.
 
 ## Citation