3 papers indexed
Dain Kim, Jinseo Kim, Sungyong Baik
Text-to-image diffusion models achieve impressive visual quality, yet demographic bias remains a challenge, as neutral prompts consistently produce stereotypical representations across gender and race. Existing approaches remain limited by costly retraining or by inference-time i…
Seongmin Kim, Kyucheol Shin, Heesun Jung, Jinseo Kim, Sungyong Baik
Text-to-image personalization aims to generate a user-provided subject in novel scenes described by text. However, most existing methods encode subject identity (fidelity) and context (editability) through the same conditioning pathway, forcing the two to compete for attention-ma…
Jinwook Jung, Taegyu Kim, Kumju Jo, Sungyong Baik
Multi-task model merging aims to consolidate several task-specific experts into a unified model, yet static merging consistently suffers from parameter interference. While dynamic merging models aim to bridge this gap, many works rely on the costly storage and loading of redundan…