Zhigang Yan

DeepSeek-AI, Daya Guo, Dejian Yang et al. · stanford, tsinghua

We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zero naturally emerges with numerous powerful and intriguing reasoning behaviors. However, it encounters challenges such as poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates multi-stage training and cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217 on reasoning tasks. To support the research community, we open-source DeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B, 70B) distilled from DeepSeek-R1 based on Qwen and Llama.

DeepSeek-AI, Aixin Liu, Aoxue Mei et al.

We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. The key technical breakthroughs of DeepSeek-V3.2 are as follows: (1) DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance in long-context scenarios. (2) Scalable Reinforcement Learning Framework: By implementing a robust reinforcement learning protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro, achieving gold-medal performance in both the 2025 International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI). (3) Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This methodology facilitates scalable agentic post-training, yielding substantial improvements in generalization and instruction-following robustness within complex, interactive environments.

Zhigang Yan, Dong Li, Zhichao Zhang et al.

In federated learning (FL), a number of devices train their local models and upload the corresponding parameters or gradients to the base station (BS) to update the global model while protecting their data privacy. However, due to the limited computation and communication resources, the number of local trainings (a.k.a. local update) and that of aggregations (a.k.a. global update) need to be carefully chosen. In this paper, we investigate and analyze the optimal trade-off between the number of local trainings and that of global aggregations to speed up the convergence and enhance the prediction accuracy over the existing works. Our goal is to minimize the global loss function under both the delay and the energy consumption constraints. In order to make the optimization problem tractable, we derive a new and tight upper bound on the loss function, which allows us to obtain closed-form expressions for the number of local trainings and that of global aggregations. Simulation results show that our proposed scheme can achieve a better performance in terms of the prediction accuracy, and converge much faster than the baseline schemes.

Zhigang Yan, Dong Li

Federated learning (FL) can lead to significant communication overhead and reliance on a central server. To address these challenges, decentralized federated learning (DFL) has been proposed as a more resilient framework. DFL involves parameter exchange between devices through a wireless network. This study analyzes the performance of resource-constrained DFL using different communication schemes (digital and analog) over wireless networks to optimize communication efficiency. Specifically, we provide convergence bounds for both digital and analog transmission approaches, enabling analysis of the model performance trained on DFL. Furthermore, for digital transmission, we investigate and analyze resource allocation between computation and communication and convergence rates, obtaining its communication complexity and the minimum probability of correction communication required for convergence guarantee. For analog transmission, we discuss the impact of channel fading and noise on the model performance and the maximum errors accumulation with convergence guarantee over fading channels. Finally, we conduct numerical simulations to evaluate the performance and convergence rate of convolutional neural networks (CNNs) and Vision Transformer (ViT) trained in the DFL framework on fashion-MNIST and CIFAR-10 datasets. Our simulation results validate our analysis and discussion, revealing how to improve performance by optimizing system parameters under different communication conditions.

DeepSeek-AI, Aixin Liu, Bei Feng et al. · stanford, tsinghua

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.

Zhigang Yan, Dong Li

In Federated Learning (FL), with parameter aggregated by a central node, the communication overhead is a substantial concern. To circumvent this limitation and alleviate the single point of failure within the FL framework, recent studies have introduced Decentralized Federated Learning (DFL) as a viable alternative. Considering the device heterogeneity, and energy cost associated with parameter aggregation, in this paper, the problem on how to efficiently leverage the limited resources available to enhance the model performance is investigated. Specifically, we formulate a problem that minimizes the loss function of DFL while considering energy and latency constraints. The proposed solution involves optimizing the number of local training rounds across diverse devices with varying resource budgets. To make this problem tractable, we first analyze the convergence of DFL with edge devices with different rounds of local training. The derived convergence bound reveals the impact of the rounds of local training on the model performance. Then, based on the derived bound, the closed-form solutions of rounds of local training in different devices are obtained. Meanwhile, since the solutions require the energy cost of aggregation as low as possible, we modify different graph-based aggregation schemes to solve this energy consumption minimization problem, which can be applied to different communication scenarios. Finally, a DFL framework which jointly considers the optimized rounds of local training and the energy-saving aggregation scheme is proposed. Simulation results show that, the proposed algorithm achieves a better performance than the conventional schemes with fixed rounds of local training, and consumes less energy than other traditional aggregation schemes.

Xiran Zhou, Yi Wen, Honghao Li et al.

Maps are fundamental medium to visualize and represent the real word in a simple and 16 philosophical way. The emergence of the 3rd wave information has made a proportion of maps are available to be generated ubiquitously, which would significantly enrich the dimensions and perspectives to understand the characteristics of the real world. However, a majority of map dataset have never been discovered, acquired and effectively used, and the map data used in many applications might not be completely fitted for the authentic demands of these applications. This challenge is emerged due to the lack of numerous well-labelled benchmark datasets for implementing the deep learning approaches into identifying complicated map content. Thus, we develop a large-scale benchmark dataset that includes well-labelled dataset for map text annotation recognition, map scene classification, map super-resolution reconstruction, and map style transferring. Furthermore, these well-labelled datasets would facilitate the state-of-the-art machine intelligence technologies to conduct map feature detection, map pattern recognition and map content retrieval. We hope our efforts would be useful for AI-enhanced cartographical applications.