RLM env stop condition fix

verifiers/envs/experimental/rlm_env.py

@@ -2658,6 +2658,13 @@ async def get_prompt_messages(self, state: State) -> Messages:
             # Subsequent turns: use parent implementation
             return await super().get_prompt_messages(state)
 
+    async def env_response(self, messages: Messages, state: State) -> Messages:                                                                                                                 
+      """Override to set final_env_response when answer is ready to avoid extra model call"""                                                                                                                          
+      tool_messages = await super().env_response(messages, state)                                                                                                                             
+      if "final_answer" in state:                                                                                                                       
+          state["final_env_response"] = tool_messages                                                                                                                                         
+      return tool_messages
+
     # =========================================================================
     # Stop Conditions
     # =========================================================================