Correction Training for Diffusion Models

arXiv:2610.14136 · 2026-10-21 공개 · arXiv · PDF

diffusion-models model-distillation denoising-steps classifier-free-guidance sample-quality exposure-bias denoiser unconditional-generation

Abstract

Iterative sampling in diffusion models can lead to error accumulation due to exposure bias. The denoiser is trained on real images interpolated with noise, but at inference its inputs are constructed from its own imperfect predictions from the earlier steps, something it has never seen at training time. This might be one reason diffusion models often struggle without additional help, such as CFG. We introduce a correction step for diffusion models which attempts to reduce this training-inference gap. The corrector takes the base model's prediction as input and is trained against the same ground-truth target. This makes correction an explicit learning task, with training inputs provided by the base model. During sampling, at each denoising step, we iteratively refine the denoiser's clean-sample prediction. This allows us to correct predictions at intermediate denoising steps, preventing early artifacts from persisting into the final sample. Although the corrector is trained only on predictions from the base denoiser, it can be applied repeatedly to its own outputs at inference for better results. This simple framework leads to substantial improvement in sample quality while largely preserving diversity. We evaluate our method across DiT, SiT, and JiT. Our method substantially improves upon previously reported FID scores for unguided generation and, notably, unconditional generation, where producing high-quality samples without guidance has been particularly challenging. Our method achieves performance competitive with classifier-free guidance (CFG), and combining correction with CFG further improves diversity and reduces oversaturation typical of CFG. Finally, we show that the additional inference cost of correction can be amortized by distilling the refined predictions back into the base diffusion model, retaining the quality gains without additional inference-time overhead.

한국어 요약

한 줄 요약

확산 모델의 예측 오류 누적 문제를 해결하기 위해 Corrector 모듈을 도입하여 샘플 품질을 향상시킨다.

핵심 기여도

핵심 아이디어

기존 확산 모델에서 Denoiser는 노이즈가 섞인 실제 이미지로 훈련되지만, 추론 시에는 이전 단계의 불완전한 예측에 기반한 입력을 받는다. 이는 Training-Inference Gap을 유발하며, 오류가 누적되어 최종 샘플에 예측 불가능한 아티팩트를 남기는 원인이다. 이를 해결하기 위해, 본 연구는 Denoiser의 예측을 입력으로 받아 동일한 Ground-Truth를 대상으로 훈련하는 Corrector 모듈을 도입한다. Corrector는 단일 단계에서 훈련되지만, 추론 시에는 반복적으로 자기 자신의 출력을 보정할 수 있어 샘플 품질을 향상시킨다. 이는 Denoiser의 오류 누적을 보완하고, Training-Inference Gap을 줄이는 핵심 아이디어이다.

기술적 접근법

주요 결과

의의 및 한계

실용적 활용