In this paper, we rethink text-to-avatar generative models by proposing TeRA, a more efficient and effective framework than the previous SDS-based models and general large 3D generative models. Our approach employs a two-stage training strategy for learning a native 3D avatar generative model. Initially, we distill a decoder to derive a structured latent space from a large human reconstruction model. Subsequently, a text-controlled latent diffusion model is trained to generate photorealistic 3D human avatars within this latent space. TeRA enhances the model performance by eliminating slow iterative optimization and enables text-based partial customization through a structured 3D human representation. Experiments have proven our approach's superiority over previous text-to-avatar generative models in subjective and objective evaluation.

Generating high-resolution 3D shapes using volumetric representations such as Signed Distance Functions (SDFs) presents substantial computational and memory challenges. We introduce Direct3D-S2, a scalable 3D generation framework based on sparse volumes that achieves superior output quality with dramatically reduced training costs. Our key innovation is the Spatial Sparse Attention (SSA) mechanism, which greatly enhances the efficiency of Diffusion Transformer (DiT) computations on sparse volumetric data. SSA allows the model to effectively process large token sets within sparse volumes, significantly reducing computational overhead and achieving a 3.9$\times$ speedup in the forward pass and a 9.6$\times$ speedup in the backward pass. Our framework also includes a variational autoencoder (VAE) that maintains a consistent sparse volumetric format across input, latent, and output stages. Compared to previous methods with heterogeneous representations in 3D VAE, this unified design significantly improves training efficiency and stability. Our model is trained on public datasets, and experiments demonstrate that Direct3D-S2 not only surpasses state-of-the-art methods in generation quality and efficiency, but also enables training at 1024³ resolution using only 8 GPUs—a task typically requiring at least 32 GPUs for volumetric representations at $256^3$ resolution, thus making gigascale 3D generation both practical and accessible. Project page: https://www.neural4d.com/research-page/direct3d-s2.

Federated learning (FL) is a privacy-preserving distributed machine learning approach that enables multiple parties to collaboratively train machine learning models without sharing local data. However, compared with the models trained on independent and identically distributed (IID) data, existing methods still face significant degradation in model performance when running on non-IID data. To solve this problem, we propose a federated learning framework based on genetic algorithm and relay strategy in this paper. The framework groups clients according to their local data distribution using genetic algorithm. After obtaining the optimal grouping result, a relay strategy is used to train and aggregate models within each group. Extensive experiments on three benchmark datasets show that FedGR significantly outperforms other state-of-the-art federated learning algorithms on various image classification tasks. The source code is available at https://github.com/zyfhylyh/FedGR.

Generating high-quality 3D assets from text and images has long been challenging, primarily due to the absence of scalable 3D representations capable of capturing intricate geometry distributions. In this work, we introduce Direct3D, a native 3D generative model scalable to in-the-wild input images, without requiring a multi-view diffusion model or SDS optimization. Our approach comprises two primary components: a Direct 3D Variational Auto-Encoder (D3D-VAE) and a Direct 3D Diffusion Transformer (D3D-DiT). D3D-VAE efficiently encodes high-resolution 3D shapes into a compact and continuous latent triplane space. Notably, our method directly supervises the decoded geometry using a semi-continuous surface sampling strategy, diverging from previous methods relying on rendered images as supervision signals. D3D-DiT models the distribution of encoded 3D latents and is specifically designed to fuse positional information from the three feature maps of the triplane latent, enabling a native 3D generative model scalable to large-scale 3D datasets. Additionally, we introduce an innovative image-to-3D generation pipeline incorporating semantic and pixel-level image conditions, allowing the model to produce 3D shapes consistent with the provided conditional image input. Extensive experiments demonstrate the superiority of our large-scale pre-trained Direct3D over previous image-to-3D approaches, achieving significantly better generation quality and generalization ability, thus establishing a new state-of-the-art for 3D content creation. Project page: https://www.neural4d.com/research/direct3d.

Large-scale cropping (LSC) is one of the mostly-used operations in desynchronization attacks and can easily destroy the watermark information by deleting continuous audio slices from the watermarked audio. In this work, we propose a spread spectrum (SS) based audio watermarking scheme to resist against LSC attacks more robustly from both theoretical and empirical perspectives. Specifically, we first perform discrete wavelet transform (DWT), graph-based transform (GBT), and singular value decomposition (SVD) on the host audio signal to produce transform coefficients. Next, we embed the chaotic encrypted watermark into DWT-GBT-SVD coefficients by the SS technique. Then, we combine m-sequences with the encrypted watermark to generate the watermarking key, which can theoretically guarantee the self-restoration of the cropped watermark based on the periodicity of m-sequences. Additionally, we develop an effective sliding window (SW) strategy to extract the fragmentary watermark slices from DWT-GBT-SVD coefficients and restore the integral watermark by the watermarking key. Finally, the proposed audio watermarking scheme, named m-SW-LSC, is compared with the state-of-the-art audio watermarking methods on audio signals with different genres and lengths under various attacks with different ratios. Experimental results demonstrate that our m-SW-LSC has a superior performance in restoring the complete watermark and exhibits significantly high robustness against LSC attacks.

Although existing malicious domains detection techniques have shown great success in many real-world applications, the problem of learning from imbalanced data is rarely concerned with this day. But the actual DNS traffic is inherently imbalanced; thus how to build malicious domains detection model oriented to imbalanced data is a very important issue worthy of study. This paper proposes a novel imbalanced malicious domains detection method based on passive DNS traffic analysis, which can effectively deal with not only the between-class imbalance problem but also the within-class imbalance problem. The experiments show that this proposed method has favorable performance compared to the existing algorithms.