As shown in the picture, not only does this question mark appear after the answer is complete, but it also keeps flashing during the large model's response process in random place. It seems that this issue did not occur in the V1 version
Whether during the output process or after it's completed, there will be no garbled text