Xiaoyu Zhang

Jujie He, Jiacai Liu, Chris Yuhao Liu et al.

The success of DeepSeek-R1 underscores the significant role of reinforcement learning (RL) in enhancing the reasoning capabilities of large language models (LLMs). In this work, we present Skywork-OR1, an effective and scalable RL implementation for long Chain-of-Thought (CoT) models. Building on the DeepSeek-R1-Distill model series, our RL approach achieves notable performance gains, increasing average accuracy across AIME24, AIME25, and LiveCodeBench from 57.8% to 72.8% (+15.0%) for the 32B model and from 43.6% to 57.5% (+13.9%) for the 7B model. Our Skywork-OR1-32B model surpasses both DeepSeek-R1 and Qwen3-32B on the AIME24 and AIME25 benchmarks, while achieving comparable results on LiveCodeBench. The Skywork-OR1-7B and Skywork-OR1-Math-7B models demonstrate competitive reasoning capabilities among models of similar size. We perform comprehensive ablation studies on the core components of our training pipeline to validate their effectiveness. Additionally, we thoroughly investigate the phenomenon of entropy collapse, identify key factors affecting entropy dynamics, and demonstrate that mitigating premature entropy collapse is critical for improved test performance. To support community research, we fully open-source our model weights, training code, and training datasets.

Kaiping Cui, Xia Feng, Liangmin Wang et al.

Secure aggregation enables federated learning (FL) to perform collaborative training of clients from local gradient updates without exposing raw data. However, existing secure aggregation schemes inevitably perform an expensive fresh setup per round because each client needs to establish fresh input-independent secrets over different rounds. The latest research, Flamingo (S&P 2023), designed a share-transfer-based reusable secret key to support the server continuously performing multiple rounds of aggregation. Nevertheless, the share transfer mechanism it proposed can only be achieved with P probability, which has limited reliability. To tackle the aforementioned problems, we propose a more reliable and anonymously authenticated scheme called Chu-ko-nu for multi-round secure aggregation. Specifically, in terms of share transfer, Chu-ko-nu breaks the probability P barrier by supplementing a redistribution process of secret key components (the sum of all components is the secret key), thus ensuring the reusability of the secret key. Based on this reusable secret key, Chu-ko-nu can efficiently perform consecutive aggregation in the following rounds. Furthermore, considering the client identity authentication and privacy protection issue most approaches ignore, Chu-ko-nu introduces a zero-knowledge proof-based authentication mechanism. It can support clients anonymously participating in FL training and enables the server to authenticate clients effectively in the presence of various attacks. Rigorous security proofs and extensive experiments demonstrated that Chu-ko-nu can provide reliable and anonymously authenticated aggregation for FL with low aggregation costs, at least a 21.02% reduction compared to the state-of-the-art schemes.

Tianwen Wei, Bo Zhu, Liang Zhao et al.

In this technical report, we introduce the training methodologies implemented in the development of Skywork-MoE, a high-performance mixture-of-experts (MoE) large language model (LLM) with 146 billion parameters and 16 experts. It is initialized from the pre-existing dense checkpoints of our Skywork-13B model. We explore the comparative effectiveness of upcycling versus training from scratch initializations. Our findings suggest that the choice between these two approaches should consider both the performance of the existing dense checkpoints and the MoE training budget. We highlight two innovative techniques: gating logit normalization, which improves expert diversification, and adaptive auxiliary loss coefficients, allowing for layer-specific adjustment of auxiliary loss coefficients. Our experimental results validate the effectiveness of these methods. Leveraging these techniques and insights, we trained our upcycled Skywork-MoE on a condensed subset of our SkyPile corpus. The evaluation results demonstrate that our model delivers strong performance across a wide range of benchmarks.

Xiaoyu Zhang, Chao Chen, Yi Xie et al.

Deep Neural Network (DNN), one of the most powerful machine learning algorithms, is increasingly leveraged to overcome the bottleneck of effectively exploring and analyzing massive data to boost advanced scientific development. It is not a surprise that cloud computing providers offer the cloud-based DNN as an out-of-the-box service. Though there are some benefits from the cloud-based DNN, the interaction mechanism among two or multiple entities in the cloud inevitably induces new privacy risks. This survey presents the most recent findings of privacy attacks and defenses appeared in cloud-based neural network services. We systematically and thoroughly review privacy attacks and defenses in the pipeline of cloud-based DNN service, i.e., data manipulation, training, and prediction. In particular, a new theory, called cloud-based ML privacy game, is extracted from the recently published literature to provide a deep understanding of state-of-the-art research. Finally, the challenges and future work are presented to help researchers to continue to push forward the competitions between privacy attackers and defenders.