Liang Zhao

Zhiyu Wu, Xiaokang Chen, Zizheng Pan et al.

We present DeepSeek-VL2, an advanced series of large Mixture-of-Experts (MoE) Vision-Language Models that significantly improves upon its predecessor, DeepSeek-VL, through two key major upgrades. For the vision component, we incorporate a dynamic tiling vision encoding strategy designed for processing high-resolution images with different aspect ratios. For the language component, we leverage DeepSeekMoE models with the Multi-head Latent Attention mechanism, which compresses Key-Value cache into latent vectors, to enable efficient inference and high throughput. Trained on an improved vision-language dataset, DeepSeek-VL2 demonstrates superior capabilities across various tasks, including but not limited to visual question answering, optical character recognition, document/table/chart understanding, and visual grounding. Our model series is composed of three variants: DeepSeek-VL2-Tiny, DeepSeek-VL2-Small and DeepSeek-VL2, with 1.0B, 2.8B and 4.5B activated parameters respectively. DeepSeek-VL2 achieves competitive or state-of-the-art performance with similar or fewer activated parameters compared to existing open-source dense and MoE-based models. Codes and pre-trained models are publicly accessible at https://github.com/deepseek-ai/DeepSeek-VL2.

Xinyuan Song, Guangji Bai, Liang Zhao

Pruning is critical for scaling large language models (LLMs). Global pruning achieves strong performance but requires $\mathcal{O}(N)$ memory, which is infeasible for billion-parameter models. Local pruning reduces GPU memory usage to that of a single layer by pruning layers independently, but it neglects inter-layer dependencies and often leads to suboptimal performance in high-sparsity regimes. Unlike unstructured pruning, structured pruning produces regular sparsity patterns that align well with GPU kernels and library optimizations, making it more hardware-efficient. However, structured pruning typically relies on global pruning, since structured patterns are more prone to severe performance degradation under local optimization. To jointly achieve structured pruning and the memory efficiency of local pruning, we propose a divide-and-conquer strategy that decomposes the global pruning problem into coordinated subproblems across different modules, each of which fits within limited GPU memory. Building on this idea, we design \textbf{STRUPRUNE}, an ADMM-based framework that integrates structured sparsity into the pruning process, combining the memory efficiency of local pruning with the hardware compatibility of structured methods. We derive a closed-form analytical solution for structured pruning masks that provides an explicit rule for layer-wise sparsity allocation, and further develop an energy-based asymptotic framework yielding a softmax-form allocation scheme that simplifies optimization while adapting to heterogeneous layer importance. Experiments demonstrate that STRUPRUNE matches the perplexity of global structured pruning while reducing memory cost from $\mathcal{O}(N)$ to $\mathcal{O}(\sqrt{N})$, enabling practical deployment at the billion-parameter scale.

Jianshen Zhu, Naveed Ahmed Azam, Kazuya Haraguchi et al.

Recently, a novel two-phase framework named mol-infer for inference of chemical compounds with prescribed abstract structures and desired property values has been proposed. The framework mol-infer is primarily based on using mixed integer linear programming (MILP) to simulate the computational process of machine learning methods and describe the necessary and sufficient conditions to ensure such a chemical graph exists. The existing approaches usually first convert the chemical compounds into handcrafted feature vectors to construct prediction functions, but because of the limit on the kinds of descriptors originated from the need for tractability in the MILP formulation, the learning performances on datasets of some properties are not good enough. A lack of good learning performance can greatly lower the quality of the inferred chemical graphs, and thus improving learning performance is of great importance. On the other hand, graph neural networks (GNN) offer a promising machine learning method to directly utilize the chemical graphs as the input, and many existing GNN-based approaches to the molecular property prediction problem have shown that they can enjoy better learning performances compared to the traditional approaches that are based on feature vectors. In this study, we develop a molecular inference framework based on mol-infer, namely mol-infer-GNN, that utilizes GNN as the learning method while keeping the great flexibility originated from the two-layered model on the abstract structure of the chemical graph to be inferred. We conducted computational experiments on the QM9 dataset to show that our proposed GNN model can obtain satisfying learning performances for some properties despite its simple structure, and can infer small chemical graphs comprising up to 20 non-hydrogen atoms within reasonable computational time.

Jianshen Zhu, Mao Takekida, Naveed Ahmed Azam et al.

Traditional QSAR/QSPR and inverse QSAR/QSPR methods often assume that chemical properties are dictated by single molecules, overlooking the influence of molecular interactions and environmental factors. In this paper, we introduce a novel QSAR/QSPR framework that can capture the combined effects of multiple molecules (e.g., small molecules or polymers) and experimental conditions on property values. We design a feature function to integrate the information of multiple molecules and the environment. Specifically, for the property Flory-Huggins $χ$-parameter, which characterizes the thermodynamic properties between the solute and the solvent, and varies in temperatures, we demonstrate through computational experimental results that our approach can achieve a competitively high learning performance compared to existing works on predicting $χ$-parameter values, while inferring the solute polymers with up to 50 non-hydrogen atoms in their monomer forms in a relatively short time. A comparison study with the simulation software J-OCTA demonstrates that the polymers inferred by our methods are of high quality.

Ryota Ido, Shengjuan Cao, Jianshen Zhu et al.

A novel framework has recently been proposed for designing the molecular structure of chemical compounds with a desired chemical property using both artificial neural networks and mixed integer linear programming. In this paper, we design a new method for inferring a polymer based on the framework. For this, we introduce a new way of representing a polymer as a form of monomer and define new descriptors that feature the structure of polymers. We also use linear regression as a building block of constructing a prediction function in the framework. The results of our computational experiments reveal a set of chemical properties on polymers to which a prediction function constructed with linear regression performs well. We also observe that the proposed method can infer polymers with up to 50 non-hydrogen atoms in a monomer form.

Naveed Ahmed Azam, Jianshen Zhu, Kazuya Haraguchi et al.

A novel framework has recently been proposed for designing the molecular structure of chemical compounds with a desired chemical property using both artificial neural networks and mixed integer linear programming. In the framework, a chemical graph with a target chemical value is inferred as a feasible solution of a mixed integer linear program that represents a prediction function and other requirements on the structure of graphs. In this paper, we propose a procedure for generating other feasible solutions of the mixed integer linear program by searching the neighbor of output chemical graph in a search space. The procedure is combined in the framework as a new building block. The results of our computational experiments suggest that the proposed method can generate an additional number of new chemical graphs with up to 50 non-hydrogen atoms.

Zhuan Cheng, Gang Wu, Hao Wu et al.

The Decentralized-Consistent-Scale (DCS) Triangle defines three dimensions that illustrate the tradeoffs of the blockchain consensus mechanism. In this paper, we propose a new hybrid consensus protocol, called Deterministic Proof of Work (DPoW), which can reach high levels of scalability and consistency without significant reduction to decentralization. Our protocol introduces a Map-reduce PoW mining algorithm to perform alongside Practical Byzantine Fault Tolerance (PBFT) verification, which together allow for transactions to be confirmed immediately, largely improving scalability. In addition, the protocol is designed such that forking cannot occur, ensuring strong consistency and security against a multitude of attacks. The Map-reduce PoW mining process ensures that no single entity can control the network, guaranteeing decentralization. We analyzed the security of our protocol by evaluating the possibility of double spending attacks, and furthermore, conducted experiments which demonstrate our claims.