This paper shows that memorization in diffusion models is governed by local data coverage and derives a theoretical criterion to predict memorized samples.

This paper establishes a connection between neural network approximation of score functions and approximation of probability distributions generated by reverse diffusion models.

The paper proposes a unified, constrained optimization framework using KL divergence and likelihood constraints to achieve effective and principled unlearning in diffusion models.

This paper derives the Oracle Distance theorem for discrete diffusion models and proves that the negative ELBO is equal to the data entropy plus the path KL from the oracle reverse process to the lear…

This paper investigates the transparency of DiffusionGemma, a model with a larger fraction of computation in a continuous latent space, and shows that it can be made more transparent by mapping inform…

The paper demonstrates that off-the-shelf image diffusion models, like Stable Diffusion, can be repurposed to generate synthetic structured data, posing a threat of ground truth drift in closed eviden…

The Parameterized Diffusion Policy (PDP) framework transforms diffusion models from general stochastic generators into precise, steerable tools for learning and adapting complex robotic behaviors by e…

This paper investigates the phenomenon of 'copying' in Distribution Matching Distillation (DMD), finding that high-dimensional distillation causes student models to spontaneously reproduce the teacher…

The paper introduces Strong Stochastic Flow Maps (SSFMs), a novel framework that directly learns the strong solution map of additive-noise Stochastic Differential Equations (SDEs), enabling few-step s…

The paper introduces D3IM, a novel parameter-free sampler that enables direct revision of visible tokens in Masked Diffusion Language Models, and proposes SCOPE to mitigate the model's tendency to per…

Equilibrated Diffusion introduces a frequency-aware approach to image customization, disentangling style and subject content embeddings to achieve superior subject fidelity and text adherence.

This paper shows that small forward-marginal error in score matching does not guarantee numerical stability, and constructs a smooth score field with small forward-marginal error but diverging moments…

The paper proposes TDPM, a time-aware diffusion model for generative recommendation, which significantly improves recommendation accuracy by explicitly modeling the non-stationary, time-evolving natur…

The paper introduces SARDI, a novel, training-free framework that uses low-confidence 'lookahead' tokens generated during the denoising process of discrete diffusion language models to dynamically gui…

This paper investigates the application of transformer-based diffusion models for simulating and reconstructing hydrological time series using data from six sites in North-East France.

This paper evaluates the consistency and effectiveness of decoding methods for diffusion large language models (dLLMs) across diverse evaluation settings and reveals their sensitivity to prompt templa…

This paper introduces a stochastic differential equation approximation for linear Temporal Difference (TD) learning under Markovian noise, explaining the constant-stepsize error floor.

This paper develops a unified spectral analysis framework to explain how knowledge transfer (KT) works across different machine learning regimes, such as Knowledge Distillation and Weak-to-Strong gene…

This paper addresses the problem of inferring a directed network from nodal measurements using graph convolutional filters and identifies the diffusion filter and network topology.