{"ID":2826234,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.19311","arxiv_id":"2512.19311","title":"MixFlow Training: Alleviating Exposure Bias with Slowed Interpolation Mixture","abstract":"This paper studies the training-testing discrepancy (a.k.a. exposure bias) problem for improving the diffusion models. During training, the input of a prediction network at one training timestep is the corresponding ground-truth noisy data that is an interpolation of the noise and the data, and during testing, the input is the generated noisy data. We present a novel training approach, named MixFlow, for improving the performance. Our approach is motivated by the Slow Flow phenomenon: the ground-truth interpolation that is the nearest to the generated noisy data at a given sampling timestep is observed to correspond to a higher-noise timestep (termed slowed timestep), i.e., the corresponding ground-truth timestep is slower than the sampling timestep. MixFlow leverages the interpolations at the slowed timesteps, named slowed interpolation mixture, for post-training the prediction network for each training timestep. Experiments over class-conditional image generation (including SiT, REPA, and RAE) and text-to-image generation validate the effectiveness of our approach. Our approach MixFlow over the RAE models achieve strong generation results on ImageNet: 1.43 FID (without guidance) and 1.10 (with guidance) at 256 x 256, and 1.55 FID (without guidance) and 1.10 (with guidance) at 512 x 512.","short_abstract":"This paper studies the training-testing discrepancy (a.k.a. exposure bias) problem for improving the diffusion models. During training, the input of a prediction network at one training timestep is the corresponding ground-truth noisy data that is an interpolation of the noise and the data, and during testing, the inpu...","url_abs":"https://arxiv.org/abs/2512.19311","url_pdf":"https://arxiv.org/pdf/2512.19311v1","authors":"[\"Hui Li\",\"Jiayue Lyu\",\"Fu-Yun Wang\",\"Kaihui Cheng\",\"Siyu Zhu\",\"Jingdong Wang\"]","published":"2025-12-22T12:00:12Z","proceeding":"cs.CV","tasks":"[\"cs.CV\",\"cs.AI\"]","methods":"[\"Diffusion Model\"]","has_code":false}