Abstract: Current indoor scene generation algorithms face significant limitations in alignment with user instructions and ensuring logical scene coherence. To address these challenges, we propose ...