Generating co-speech gestures in real time requires both temporal coherence and efficient sampling. We introduce Accelerated Rolling Diffusion, a novel framework for streaming gesture generation that extends rolling diffusion models with structured progressive noise scheduling, enabling seamless long-sequence motion synthesis while preserving realism and diversity. We further propose Rolling Diffusion Ladder Acceleration (RDLA), a new approach that restructures the noise schedule into a stepwise ladder, allowing multiple frames to be denoised simultaneously. This significantly improves sampling efficiency while maintaining motion consistency, achieving up to a 2x speedup with high visual fidelity and temporal coherence. We evaluate our approach on ZEGGS and BEAT, strong benchmarks for real-world applicability. Our framework is universally applicable to any diffusion-based gesture generation model, transforming it into a streaming approach. Applied to three state-of-the-art methods, it consistently outperforms them, demonstrating its effectiveness as a generalizable and efficient solution for real-time, high-fidelity co-speech gesture synthesis.

The Latent Stochastic Differential Equation (SDE) is a powerful tool for time series and sequence modeling. However, training Latent SDEs typically relies on adjoint sensitivity methods, which depend on simulation and backpropagation through approximate SDE solutions, which limit scalability. In this work, we propose SDE Matching, a new simulation-free method for training Latent SDEs. Inspired by modern Score- and Flow Matching algorithms for learning generative dynamics, we extend these ideas to the domain of stochastic dynamics for time series and sequence modeling, eliminating the need for costly numerical simulations. Our results demonstrate that SDE Matching achieves performance comparable to adjoint sensitivity methods while drastically reducing computational complexity.

It is generally accepted that starting neural networks training with large learning rates (LRs) improves generalization. Following a line of research devoted to understanding this effect, we conduct an empirical study in a controlled setting focusing on two questions: 1) how large an initial LR is required for obtaining optimal quality, and 2) what are the key differences between models trained with different LRs? We discover that only a narrow range of initial LRs slightly above the convergence threshold lead to optimal results after fine-tuning with a small LR or weight averaging. By studying the local geometry of reached minima, we observe that using LRs from this optimal range allows for the optimization to locate a basin that only contains high-quality minima. Additionally, we show that these initial LRs result in a sparse set of learned features, with a clear focus on those most relevant for the task. In contrast, starting training with too small LRs leads to unstable minima and attempts to learn all features simultaneously, resulting in poor generalization. Conversely, using initial LRs that are too large fails to detect a basin with good solutions and extract meaningful patterns from the data.

Diffusion distillation methods aim to compress the diffusion models into efficient one-step generators while trying to preserve quality. Among them, Distribution Matching Distillation (DMD) offers a suitable framework for training general-form one-step generators, applicable beyond unconditional generation. In this work, we introduce its modification, called Regularized Distribution Matching Distillation, applicable to unpaired image-to-image (I2I) problems. We demonstrate its empirical performance in application to several translation tasks, including 2D examples and I2I between different image datasets, where it performs on par or better than multi-step diffusion baselines.

Generative Flow Networks (GFlowNets) treat sampling from distributions over compositional discrete spaces as a sequential decision-making problem, training a stochastic policy to construct objects step by step. Recent studies have revealed strong connections between GFlowNets and entropy-regularized reinforcement learning. Building on these insights, we propose to enhance planning capabilities of GFlowNets by applying Monte Carlo Tree Search (MCTS). Specifically, we show how the MENTS algorithm (Xiao et al., 2019) can be adapted for GFlowNets and used during both training and inference. Our experiments demonstrate that this approach improves the sample efficiency of GFlowNet training and the generation fidelity of pre-trained GFlowNet models.

Conventional diffusion models often rely on a fixed forward process, which implicitly defines complex marginal distributions over latent variables. This can often complicate the reverse process' task in learning generative trajectories, and results in costly inference for diffusion models. To address these limitations, we introduce Neural Flow Diffusion Models (NFDM), a novel framework that enhances diffusion models by supporting a broader range of forward processes beyond the standard linear Gaussian. We also propose a novel parameterization technique for learning the forward process. Our framework provides an end-to-end, simulation-free optimization objective, effectively minimizing a variational upper bound on the negative log-likelihood. Experimental results demonstrate NFDM's strong performance, evidenced by state-of-the-art likelihoods across a range of image generation tasks. Furthermore, we investigate NFDM's capacity for learning generative dynamics with specific characteristics, such as deterministic straight lines trajectories, and demonstrate how the framework can be adopted for learning bridges between two distributions. The results underscores NFDM's versatility and its potential for a wide range of applications.

Protein design requires a deep understanding of the inherent complexities of the protein universe. While many efforts lean towards conditional generation or focus on specific families of proteins, the foundational task of unconditional generation remains underexplored and undervalued. Here, we explore this pivotal domain, introducing DiMA, a model that leverages continuous diffusion on embeddings derived from the protein language model, ESM-2, to generate amino acid sequences. DiMA surpasses leading solutions, including autoregressive transformer-based and discrete diffusion models, and we quantitatively illustrate the impact of the design choices that lead to its superior performance. We extensively evaluate the quality, diversity, distribution similarity, and biological relevance of the generated sequences using multiple metrics across various modalities. Our approach consistently produces novel, diverse protein sequences that accurately reflect the inherent structural and functional diversity of the protein space. This work advances the field of protein design and sets the stage for conditional models by providing a robust framework for scalable and high-quality protein sequence generation.

Drawing inspiration from the success of diffusion models in various domains, numerous research papers proposed methods for adapting them to text data. Despite these efforts, none of them has managed to achieve the quality of the large language models. In this paper, we conduct a comprehensive analysis of key components of the text diffusion models and introduce a novel approach named Text Encoding Diffusion Model (TEncDM). Instead of the commonly used token embedding space, we train our model in the space of the language model encodings. Additionally, we propose to use a Transformer-based decoder that utilizes contextual information for text reconstruction. We also analyse self-conditioning and find that it increases the magnitude of the model outputs, allowing the reduction of the number of denoising steps at the inference stage. Evaluation of TEncDM on two downstream text generation tasks, QQP and XSum, demonstrates its superiority over existing non-autoregressive models.

Generative adversarial networks have recently demonstrated outstanding performance in neural vocoding outperforming best autoregressive and flow-based models. In this paper, we show that this success can be extended to other tasks of conditional audio generation. In particular, building upon HiFi vocoders, we propose a novel HiFi++ general framework for bandwidth extension and speech enhancement. We show that with the improved generator architecture, HiFi++ performs better or comparably with the state-of-the-art in these tasks while spending significantly less computational resources. The effectiveness of our approach is validated through a series of extensive experiments.

Inspired by recent research that recommends starting neural networks training with large learning rates (LRs) to achieve the best generalization, we explore this hypothesis in detail. Our study clarifies the initial LR ranges that provide optimal results for subsequent training with a small LR or weight averaging. We find that these ranges are in fact significantly narrower than generally assumed. We conduct our main experiments in a simplified setup that allows precise control of the learning rate hyperparameter and validate our key findings in a more practical setting.