Search

In this paper, we introduce Hunyuan-Large, which is currently the largest open-source Transformer-based mixture of experts model, with a total of 389 billion parameters and 52 billion activation parameters, capable of handling up to 256K tokens. We conduct a thorough evaluation of Hunyuan-Large's superior performance across various benchmarks including language understanding and generation, logical reasoning, mathematical problem-solving, coding, long-context, and aggregated tasks, where it outperforms LLama3.1-70B and exhibits comparable performance when compared to the significantly larger LLama3.1-405B model. Key practice of Hunyuan-Large include large-scale synthetic data that is orders larger than in previous literature, a mixed expert routing strategy, a key-value cache compression technique, and an expert-specific learning rate strategy. Additionally, we also investigate the scaling laws and learning rate schedule of mixture of experts models, providing valuable insights and guidances for future model development and optimization. The code and checkpoints of Hunyuan-Large are released to facilitate future innovations and applications. Codes: https://github.com/Tencent/Hunyuan-Large Models: https://huggingface.co/tencent/Tencent-Hunyuan-Large, Comment: 17 pages, 4 Figures

The sophistication and diversity of contemporary cyberattacks have rendered the use of proxies, gateways, firewalls, and encrypted tunnels as a standalone defensive strategy inadequate. Consequently, the proactive identification of data anomalies has emerged as a prominent area of research within the field of data security. The majority of extant studies concentrate on sample equilibrium data, with the consequence that the detection effect is not optimal in the context of unbalanced data. In this study, the unsupervised learning method is employed to identify anomalies in dynamic data flows. Initially, multi-dimensional features are extracted from real-time data, and a clustering algorithm is utilised to analyse the patterns of the data. This enables the potential outliers to be automatically identified. By clustering similar data, the model is able to detect data behaviour that deviates significantly from normal traffic without the need for labelled data. The results of the experiments demonstrate that the proposed method exhibits high accuracy in the detection of anomalies across a range of scenarios. Notably, it demonstrates robust and adaptable performance, particularly in the context of unbalanced data.

This paper presents a lightweight image fusion algorithm specifically designed for merging visible light and infrared images, with an emphasis on balancing performance and efficiency. The proposed method enhances the generator in a Generative Adversarial Network (GAN) by integrating the Convolutional Block Attention Module (CBAM) to improve feature focus and utilizing Depthwise Separable Convolution (DSConv) for more efficient computations. These innovations significantly reduce the model's computational cost, including the number of parameters and inference latency, while maintaining or even enhancing the quality of the fused images. Comparative experiments using the M3FD dataset demonstrate that the proposed algorithm not only outperforms similar image fusion methods in terms of fusion quality but also offers a more resource-efficient solution suitable for deployment on embedded devices. The effectiveness of the lightweight design is validated through extensive ablation studies, confirming its potential for real-time applications in complex environments.

Decomposition of skin pigment plays an important role in medical fields. Human skin can be decomposed into two primitive components, hemoglobin and melanin. It is our goal to apply these results for diagnosis of skin cancer. In this paper, various methods for skin pigment decomposition are reviewed comparatively and the performance of each method is evaluated both theoretically and experimentally. In addition, isometric feature mapping (Isomap) is introduced in order to improve the dimensionality reduction performance in context of skin pigment decomposition., Comment: 5 pages, 7 figures

Omega Centauri, the largest known globular cluster in the Milky Way, is believed to be the remains of a dwarf galaxy's core. Giving its potential abundance of dark matter (DM), it is an attractive target for investigating the nature of this elusive substance in our local environment. Our study demonstrates that by observing Omega Centauri with the SKA for 1000 hours, we can detect synchrotron radio or Inverse Compton (IC) emissions from the DM annihilation products. It enables us to constrain the cross-section of DM annihilation down to $\sim {\rm 10^{-30}~cm^3~s^{-1}}$ for DM mass from several $\rm{GeV}$ to $\rm{100~GeV}$, which is much stronger compared with other observations. Additionally, we explore the axion, another well-motivated DM candidate, and provide stimulated decay calculations. It turns out that the sensitivity can reach $g_{\rm{a\gamma\gamma}} \sim 10^{-10} ~\rm{GeV^{-1}}$ for $2\times 10^{-7} ~\rm{eV} < m_a < 2\times 10^{-4} ~\rm{eV}$., Comment: 20 pages, 8 figures

In this work, we propose a string-inspired two fields inflation model to address the fine-tuning problem that the standard inflation model suffers. The fast-rolling tachyon $\mathcal{T}$ originated from the D-brane and anti-D-brane pair annihilation locks the inflaton $\varphi$ slowly rolling on a Higgs-like potential $V(\varphi)=-m_\varphi^2\varphi^2+\lambda \varphi^4$ and drives a kinetically stabilized (KS) inflation. Our numerical simulation confirms such a solution is a dynamic attractor. In particular, for $\lambda< 0.8\times 10^{-3}$, the e-folding number contributed by the KS inflation phase can be larger than $62$ to solve the horizon and flatness problems of Big Bang theory. Notably, this KS inflation generates a nearly scale-invariant primordial curvature perturbations spectrum consistent with current cosmic microwave background (CMB) observations. It predicts a low tensor-to-scalar ratio, which the current primordial gravitational wave background (the B-modes in CMB) searches favor., Comment: 16 pages, 4 figures

Unmanned aerial vehicles (UAVs) have various advantages, but their practical applications are influenced by their limited energy. Therefore, it is important to manage their power consumption and also important to establish corresponding power consumption models. However, most of existing works either establish theoretical power consumption models for fixed-wing UAVs and single-rotor UAVs, or provide heuristic power consumption models for multi-rotor UAVs without rigorous mathematical derivations. This paper aims to establish theoretical power consumption models for multi-rotor UAVs. To be specific, the closed-form power consumption models for a multi-rotor UAV in three flight statuses, i.e., forward flight, vertical ascent and vertical descent, are derived by leveraging the relationship between single-rotor UAVs and multi-rotor UAVs in terms of power consumptions. On this basis, a generic flight power consumption model for the UAV in a three-dimensional (3-D) scenario is obtained. Extensive experiments are conducted by using DJI M210 and a mobile app made by DJI Mobile SDK in real scenarios, and confirm the correctness and effectiveness of these models; in addition, simulations are performed to further investigate the effect of the rotor numbers on the power consumption for the UAV. The proposed power consumption models not only reveal how the power consumption of multi-rotor UAVs are affected by various factors, but also pave the way for introducing other novel applications.

E-commerce platforms generate vast amounts of customer behavior data, such as clicks and purchases, from millions of unique users every day. However, effectively using this data for behavior understanding tasks is challenging because there are usually not enough labels to learn from all users in a supervised manner. This paper extends the BERT model to e-commerce user data for pre-training representations in a self-supervised manner. By viewing user actions in sequences as analogous to words in sentences, we extend the existing BERT model to user behavior data. Further, our model adopts a unified structure to simultaneously learn from long-term and short-term user behavior, as well as user attributes. We propose methods for the tokenization of different types of user behavior sequences, the generation of input representation vectors, and a novel pretext task to enable the pre-trained model to learn from its own input, eliminating the need for labeled training data. Extensive experiments demonstrate that the learned representations result in significant improvements when transferred to three different real-world tasks, particularly compared to task-specific modeling and multi-task representation learning