It seems the self.progression variable is not reset in the reset function here https://github.com/balrog-ai/BALROG/blob/main/balrog/environments/babyai_text/clean_lang_wrapper.py#L45-L54. It makes it so that once one episode has been successful all future ones will also get progression = 1.0. Is this intended? If not I'm happy to make a PR to fix it.