Update README.md

README.md

@@ -141,7 +141,7 @@ Hindi:
 モデルが理解できない概念ではハルシネーション出現しやすくなります  
 アウトプット利用は自己責任でお願いします
 
-The model tends to exhibit unstable stop token behavior and repeated translations in short sentences or very long texts. Performance issues, such as skipping parts of the text, become noticeable in translations exceeding 4000 tokens. Additionally, the model is prone to hallucinations when dealing with concepts it cannot understand. Please use the output at your own risk and cooperate in mitigating the risks associated with the model's release.
+In short sentences or very long texts, the model may experience unstable stop token behavior and duplicate translations. When translating texts longer than 4000 tokens, performance degradation, such as skipping content, becomes evident. The model is also more likely to produce hallucinations when encountering concepts it does not understand. Please use the output at your own risk.
 
 ### Todo  
 For the support of multiple languages, there are areas where accuracy has been compromised. We will continue to train more accurate models.