Add exception for conversation + q/a and add breaking change warning …

…to changelog

src/promptflow-evals/CHANGELOG.md

@@ -9,6 +9,9 @@
 ### Bugs Fixed
 - Fixed evaluators to accept (non-Azure) Open AI Configs.
 
+### Breaking Changes
+- Replaced `jailbreak` parameter in `AdversarialSimulator` with `_jailbreak_type` parameter to support multiple jailbreak types. Instead of editing this parameter directly, we recommend using the `JailbreakAdversarialSimulator` class for UPIA jailbreak and `IndirectAttackSimulator` class for XPIA jailbreak.
+
 ### Improvements
 - Set the PF_EVALS_BATCH_USE_ASYNC environment variable to True by default to enable asynchronous batch run for async-enabled built-in evaluators, improving performance.
 

src/promptflow-evals/promptflow/evals/evaluators/_xpia/_xpia.py

@@ -113,6 +113,9 @@ def __call__(
         :return: The evaluation scores and reasoning.
         :rtype: dict
         """
+        if conversation and any([question, answer]):
+            raise ValueError("'conversation' cannot be provided if 'question' and 'answer' are also provided.")
+
         if conversation:
             self._validate_conversation(conversation)