EPISODE · Oct 16, 2025 · 20 MIN
Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models
from Daily Paper Cast · host Jingwen Liang, Gengyu Wang
🤗 Upvotes: 26 | cs.LG, cs.AI Authors: Youngrok Park, Hojung Jung, Sangmin Bae, Se-Young Yun Title: Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models Arxiv: http://arxiv.org/abs/2510.11057v1 Abstract: Diffusion models have achieved remarkable success as generative models. However, even a well-trained model can accumulate errors throughout the generation process. These errors become particularly problematic when arbitrary guidance is applied to steer samples toward desired properties, which often breaks sample fidelity. In this paper, we propose a general solution to address the off-manifold phenomenon observed in diffusion models. Our approach leverages a time predictor to estimate deviations from the desired data manifold at each timestep, identifying that a larger time gap is associated with reduced generation quality. We then design a novel guidance mechanism, `Temporal Alignment Guidance' (TAG), attracting the samples back to the desired manifold at every timestep during generation. Through extensive experiments, we demonstrate that TAG consistently produces samples closely aligned with the desired manifold at each timestep, leading to significant improvements in generation quality across various downstream tasks.
What this episode covers
🤗 Upvotes: 26 | cs.LG, cs.AI Authors: Youngrok Park, Hojung Jung, Sangmin Bae, Se-Young Yun Title: Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models Arxiv: http://arxiv.org/abs/2510.11057v1 Abstract: Diffusion models have achieved remarkable success as generative models. However, even a well-trained model can accumulate errors throughout the generation process. These errors become particularly problematic when arbitrary guidance is applied to steer samples toward desired properties, which often breaks sample fidelity. In this paper, we propose a general solution to address the off-manifold phenomenon observed in diffusion models. Our approach leverages a time predictor to estimate deviations from the desired data manifold at each timestep, identifying that a larger time gap is associated with reduced generation quality. We then design a novel guidance mechanism, `Temporal Alignment Guidance' (TAG), attracting the samples back to the desired manifold at every timestep during generation. Through extensive experiments, we demonstrate that TAG consistently produces samples closely aligned with the desired manifold at each timestep, leading to significant improvements in generation quality across various downstream tasks.
NOW PLAYING
Temporal Alignment Guidance: On-Manifold Sampling in Diffusion Models
No transcript for this episode yet
Similar Episodes
No similar episodes found.