No other representation component is needed: Diffusion transformers can provide representation guidance by themselves

Rating
4 - Good
Authors
Dengyang Jiang, Mengmeng Wang
Date
2025
Review Status
Todo

Summary

•
Method
◦
external representation model 없이, EMA 모델의 higher layer feature를 target feature alignment으로 사용하여 훈련
•
Result
◦
초반 수렴은 REPA가 빠르고, 최종 수렴에서는 유사한 성능 달성