Running the code revealed some issues
The role of the scene graph seems not particularly evident. In most cases, text navigation degrades to object navigation, which goes against the original intention of text navigation. Especially for some successful text navigation, it is actually the provided infos['goal_name'] that plays a role. Is my conjecture correct?I hope to get your reply，thank you!
Running the code revealed some issues
The role of the scene graph seems not particularly evident. In most cases, text navigation degrades to object navigation, which goes against the original intention of text navigation. Especially for some successful text navigation, it is actually the provided infos['goal_name'] that plays a role. Is my conjecture correct?I hope to get your reply，thank you!