@misc{indiciae43eb55a81b36, title = {Structural Self-Teaching for Compositional Generation in Unified Multimodal Models}, author = {Jaeyoo Park and Bohyung Han}, year = {2026}, url = {https://searcharxiv.com/abs/2611.18261}, note = {Source identifier: 2611.18261} }