update README: vmlu test

README.md

@@ -102,9 +102,15 @@ We evaluated our model via peer comparison on multiple publicly available datase
 
 Based on this results, our model performs on-par or better than most models for tasks in Vietnamese and demonstrate that this approach is extremely potential.
 
+While this model primarily specializes in multi-turn conversational scenarios, it has demonstrated its competence in various multiple-choice question and answer tasks during testing. Below, you can find the results, fairly evaluated by the [VMLU team](https://vmlu.ai), in comparison to other open-source models, including VBD-LLaMA2-7B-50b-Chat. (We extend our gratitude to the VMLU team for their diligent work in creating an open-source public evaluation dataset).
+
+<p align="center"> <img src="vmlu.png" width="500" /> </p>
+
+<p align="center"> Table 3. <a href="https://vmlu.ai/leaderboard"> Benchmark on VMLU datasets </a> </p>
+
 Pretraining loss:
 
-<p align="left"> <img src="loss.png" width="500" /> </p>
+<p align="center"> <img src="loss.png" width="500" /> </p>
 
 <h3> Run the model </h3>