Update README.md

Replace the scores to ones with batch size 1

README.md

@@ -48,13 +48,13 @@ In addition, Sarashina2.2-3B outperforms Sarashina2-70B in Japanese math and cod
 
 #### Evaluation in Japanese tasks
 
-| Model            | NIILC      | JMMLU      | MGSM-ja   | JHumanEval |
-|------------------|------------|------------|-----------|------------|
-| [Sarashina2-7B](https://huggingface.co/sbintuitions/sarashina2-7b)    | 62.2       | 42.5       | 7.2       | 12.8       |
-| [Sarashina2-70B](https://huggingface.co/sbintuitions/sarashina2-70b)   | **66.1**   | **62.7**   | 56.4      | 22.0       |
-|**[Sarashina2.2-0.5B](https://huggingface.co/sbintuitions/sarashina2.2-0.5b)**| 34.6    | 28.8       | 21.2      | 15.2       |
-|**[Sarashina2.2-1B](https://huggingface.co/sbintuitions/sarashina2.2-1b)**| 47.2      | 38.4       | 38.8      | 21.3       |
-|**[Sarashina2.2-3B](https://huggingface.co/sbintuitions/sarashina2.2-3b)**| 62.2      | 52.7       | **63.6**  | **39.6**   |
+| Model                                                                          | NIILC      | JMMLU     | MGSM-ja    | JHumanEval  |
+|--------------------------------------------------------------------------------|------------|-----------|------------|-------------|
+| [Sarashina2-7B](https://huggingface.co/sbintuitions/sarashina2-7b)             | 61.4       | 42.5      | 8.4        | 12.8        |
+| [Sarashina2-70B](https://huggingface.co/sbintuitions/sarashina2-70b)           | **65.4**   | **62.7**  | 54.0       | 22.0        |
+| **[Sarashina2.2-0.5B](https://huggingface.co/sbintuitions/sarashina2.2-0.5b)** | 33.9       | 28.8      | 21.6       | 15.2        |
+| **[Sarashina2.2-1B](https://huggingface.co/sbintuitions/sarashina2.2-1b)**     | 47.2       | 38.2      | 39.6       | 20.7        |
+| **[Sarashina2.2-3B](https://huggingface.co/sbintuitions/sarashina2.2-3b)**     | 63.0       | 52.7      | **63.6**   | **39.0**    |
 
 
 ## Ethical Considerations and Limitations