Update README.md

README.md

@@ -71,7 +71,7 @@ We observe that across all models, the performance gain is logarithmic in relati
 How to use bloomz-560m-sft-chat
 -------------------------------
 
-There are no specific instructions for using these models in a normal causal inference context. However, to leverage the chatbot capability of the model, an individual's prompt should be preceded by the EOS token ({</s>}), and the generated part should be preceded by the BOS token ({<s>}). The structure takes the following form:
+There are no specific instructions for using these models in a normal causal inference context. However, to leverage the chatbot capability of the model, an individual's prompt should be preceded by the EOS token (&lt;/s>), and the generated part should be preceded by the BOS token (&lt;s>). The structure takes the following form:
 ```
 </s>[human prompt 1]<s>[bot answer 1]</s>[human prompt 2]<s>
 ```