method · first seen 13 Nov 2026
structural self-teaching
Training unified multimodal diffusion models using the understanding path's grounding masks as structural supervision for compositional image generation.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Structural Self-Teaching for Compositional Generation in Unified Multimodal ModelsPapers · 13 Nov 2026 · Jaeyoo Park, Bohyung Han
Latest sources
- Structural Self-Teaching for Compositional Generation in Unified Multimodal ModelsPapers · 13 Nov 2026 · Jaeyoo Park, Bohyung Han