We sincerely appreciate your efforts and open-source contributions. Would you consider releasing the SD3.5-based MixFlow? It would be of great help to the diffusion model community.
Additionally, at least four recent studies on exposure bias in diffusion models (see [1-4]) were published in 2025 but are missing from your Related Work and experiments. We suggest adding them to further enhance the paper's impact.
Finally, please note a citation error regarding the paper "Manifold constraint reduces exposure bias in accelerated diffusion sample": it was published in ICLR 2025, but your reference (https://arxiv.org/pdf/2512.19311) lists it as ICLR 2024.
[1] On error propagation of diffusion models (ICLR 2024)
[2] Improved diffusion-based generative model with better adversarial robustness (ICLR 2025)
[3] Anti-exposure bias in diffusion models (ICLR 2025)
[4] Frequency regulation for exposure bias mitigation in diffusion models (ACM MM 2025)
We sincerely appreciate your efforts and open-source contributions. Would you consider releasing the SD3.5-based MixFlow? It would be of great help to the diffusion model community.
Additionally, at least four recent studies on exposure bias in diffusion models (see [1-4]) were published in 2025 but are missing from your Related Work and experiments. We suggest adding them to further enhance the paper's impact.
Finally, please note a citation error regarding the paper "Manifold constraint reduces exposure bias in accelerated diffusion sample": it was published in ICLR 2025, but your reference (https://arxiv.org/pdf/2512.19311) lists it as ICLR 2024.
[1] On error propagation of diffusion models (ICLR 2024)
[2] Improved diffusion-based generative model with better adversarial robustness (ICLR 2025)
[3] Anti-exposure bias in diffusion models (ICLR 2025)
[4] Frequency regulation for exposure bias mitigation in diffusion models (ACM MM 2025)