Na Wang

Jintao Guo, Na Wang, Lei Qi et al.

Domain generalization (DG) aims to learn a model that generalizes well to unseen target domains utilizing multiple source domains without re-training. Most existing DG works are based on convolutional neural networks (CNNs). However, the local operation of the convolution kernel makes the model focus too much on local representations (e.g., texture), which inherently causes the model more prone to overfit to the source domains and hampers its generalization ability. Recently, several MLP-based methods have achieved promising results in supervised learning tasks by learning global interactions among different patches of the image. Inspired by this, in this paper, we first analyze the difference between CNN and MLP methods in DG and find that MLP methods exhibit a better generalization ability because they can better capture the global representations (e.g., structure) than CNN methods. Then, based on a recent lightweight MLP method, we obtain a strong baseline that outperforms most state-of-the-art CNN-based methods. The baseline can learn global structure representations with a filter to suppress structure irrelevant information in the frequency space. Moreover, we propose a dynAmic LOw-Frequency spectrum Transform (ALOFT) that can perturb local texture features while preserving global structure features, thus enabling the filter to remove structure-irrelevant information sufficiently. Extensive experiments on four benchmarks have demonstrated that our method can achieve great performance improvement with a small number of parameters compared to SOTA CNN-based DG methods. Our code is available at https://github.com/lingeringlight/ALOFT/.

Huichi Zhou, Siyuan Guo, Anjie Liu et al.

We introduce \emph{Memento-Skills}, a generalist, continually-learnable LLM agent system that functions as an \emph{agent-designing agent}: it autonomously constructs, adapts, and improves task-specific agents through experience. The system is built on a memory-based reinforcement learning framework with \emph{stateful prompts}, where reusable skills (stored as structured markdown files) serve as persistent, evolving memory. These skills encode both behaviour and context, enabling the agent to carry forward knowledge across interactions. Starting from simple elementary skills (like Web search and terminal operations), the agent continually improves via the \emph{Read--Write Reflective Learning} mechanism introduced in \emph{Memento~2}~\cite{wang2025memento2}. In the \emph{read} phase, a behaviour-trainable skill router selects the most relevant skill conditioned on the current stateful prompt; in the \emph{write} phase, the agent updates and expands its skill library based on new experience. This closed-loop design enables \emph{continual learning without updating LLM parameters}, as all adaptation is realised through the evolution of externalised skills and prompts. Unlike prior approaches that rely on human-designed agents, Memento-Skills enables a generalist agent to \emph{design agents end-to-end} for new tasks. Through iterative skill generation and refinement, the system progressively improves its own capabilities. Experiments on the \emph{General AI Assistants} benchmark and \emph{Humanity's Last Exam} demonstrate sustained gains, achieving 26.2\% and 116.2\% relative improvements in overall accuracy, respectively. Code is available at https://github.com/Memento-Teams/Memento-Skills.

Gongning Luo, Kuanquan Wang, Jun Liu et al.

Efficient automatic segmentation of multi-level (i.e. main and branch) pulmonary arteries (PA) in CTPA images plays a significant role in clinical applications. However, most existing methods concentrate only on main PA or branch PA segmentation separately and ignore segmentation efficiency. Besides, there is no public large-scale dataset focused on PA segmentation, which makes it highly challenging to compare the different methods. To benchmark multi-level PA segmentation algorithms, we organized the first \textbf{P}ulmonary \textbf{AR}tery \textbf{SE}gmentation (PARSE) challenge. On the one hand, we focus on both the main PA and the branch PA segmentation. On the other hand, for better clinical application, we assign the same score weight to segmentation efficiency (mainly running time and GPU memory consumption during inference) while ensuring PA segmentation accuracy. We present a summary of the top algorithms and offer some suggestions for efficient and accurate multi-level PA automatic segmentation. We provide the PARSE challenge as open-access for the community to benchmark future algorithm developments at \url{https://parse2022.grand-challenge.org/Parse2022/}.

Ailin Huang, Ang Li, Aobo Kong et al.

We introduce Step 3.5 Flash, a sparse Mixture-of-Experts (MoE) model that bridges frontier-level agentic intelligence and computational efficiency. We focus on what matters most when building agents: sharp reasoning and fast, reliable execution. Step 3.5 Flash pairs a 196B-parameter foundation with 11B active parameters for efficient inference. It is optimized with interleaved 3:1 sliding-window/full attention and Multi-Token Prediction (MTP-3) to reduce the latency and cost of multi-round agentic interactions. To reach frontier-level intelligence, we design a scalable reinforcement learning framework that combines verifiable signals with preference feedback, while remaining stable under large-scale off-policy training, enabling consistent self-improvement across mathematics, code, and tool use. Step 3.5 Flash demonstrates strong performance across agent, coding, and math tasks, achieving 85.4% on IMO-AnswerBench, 86.4% on LiveCodeBench-v6 (2024.08-2025.05), 88.2% on tau2-Bench, 69.0% on BrowseComp (with context management), and 51.0% on Terminal-Bench 2.0, comparable to frontier models such as GPT-5.2 xHigh and Gemini 3.0 Pro. By redefining the efficiency frontier, Step 3.5 Flash provides a high-density foundation for deploying sophisticated agents in real-world industrial environments.

Guoqing Ma, Haoyang Huang, Kun Yan et al.

We present Step-Video-T2V, a state-of-the-art text-to-video pre-trained model with 30B parameters and the ability to generate videos up to 204 frames in length. A deep compression Variational Autoencoder, Video-VAE, is designed for video generation tasks, achieving 16x16 spatial and 8x temporal compression ratios, while maintaining exceptional video reconstruction quality. User prompts are encoded using two bilingual text encoders to handle both English and Chinese. A DiT with 3D full attention is trained using Flow Matching and is employed to denoise input noise into latent frames. A video-based DPO approach, Video-DPO, is applied to reduce artifacts and improve the visual quality of the generated videos. We also detail our training strategies and share key observations and insights. Step-Video-T2V's performance is evaluated on a novel video generation benchmark, Step-Video-T2V-Eval, demonstrating its state-of-the-art text-to-video quality when compared with both open-source and commercial engines. Additionally, we discuss the limitations of current diffusion-based model paradigm and outline future directions for video foundation models. We make both Step-Video-T2V and Step-Video-T2V-Eval available at https://github.com/stepfun-ai/Step-Video-T2V. The online version can be accessed from https://yuewen.cn/videos as well. Our goal is to accelerate the innovation of video foundation models and empower video content creators.

Ailin Huang, Boyong Wu, Bruce Wang et al.

Real-time speech interaction, serving as a fundamental interface for human-machine collaboration, holds immense potential. However, current open-source models face limitations such as high costs in voice data collection, weakness in dynamic control, and limited intelligence. To address these challenges, this paper introduces Step-Audio, the first production-ready open-source solution. Key contributions include: 1) a 130B-parameter unified speech-text multi-modal model that achieves unified understanding and generation, with the Step-Audio-Chat version open-sourced; 2) a generative speech data engine that establishes an affordable voice cloning framework and produces the open-sourced lightweight Step-Audio-TTS-3B model through distillation; 3) an instruction-driven fine control system enabling dynamic adjustments across dialects, emotions, singing, and RAP; 4) an enhanced cognitive architecture augmented with tool calling and role-playing abilities to manage complex tasks effectively. Based on our new StepEval-Audio-360 evaluation benchmark, Step-Audio achieves state-of-the-art performance in human evaluations, especially in terms of instruction following. On open-source benchmarks like LLaMA Question, shows 9.3% average performance improvement, demonstrating our commitment to advancing the development of open-source multi-modal language technologies. Our code and models are available at https://github.com/stepfun-ai/Step-Audio.

Qi Su, Na Wang, Jiawen Xie et al.

The automatic lung lobe segmentation algorithm is of great significance for the diagnosis and treatment of lung diseases, however, which has great challenges due to the incompleteness of pulmonary fissures in lung CT images and the large variability of pathological features. Therefore, we propose a new automatic lung lobe segmentation framework, in which we urge the model to pay attention to the area around the pulmonary fissure during the training process, which is realized by a task-specific loss function. In addition, we introduce an end-to-end pulmonary fissure generation method in the auxiliary pulmonary fissure segmentation task, without any additional network branch. Finally, we propose a registration-based loss function to alleviate the convergence difficulty of the Dice loss supervised pulmonary fissure segmentation task. We achieve 97.83% and 94.75% dice scores on our private dataset STLB and public LUNA16 dataset respectively.

Boyong Wu, Chao Yan, Chen Hu et al.

This paper presents Step-Audio 2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech conversation, Step-Audio 2 incorporates the generation of discrete audio tokens into language modeling, significantly enhancing its responsiveness to paralinguistic information such as speaking styles and emotions. To effectively leverage the rich textual and acoustic knowledge in real-world data, Step-Audio 2 integrates retrieval-augmented generation (RAG) and is able to call external tools such as web search to mitigate hallucination and audio search to switch timbres. Trained on millions of hours of speech and audio data, Step-Audio 2 delivers intelligence and expressiveness across diverse conversational scenarios. Evaluation results demonstrate that Step-Audio 2 achieves state-of-the-art performance on various audio understanding and conversational benchmarks compared to other open-source and commercial solutions. Please visit https://github.com/stepfun-ai/Step-Audio2 for more information.

Chen Hu, Haikuo Du, Heng Wang et al.

As LLMs shift toward autonomous agents, Deep Research has emerged as a pivotal metric. However, existing academic benchmarks like BrowseComp often fail to meet real-world demands for open-ended research, which requires robust skills in intent recognition, long-horizon decision-making, and cross-source verification. To address this, we introduce Step-DeepResearch, a cost-effective, end-to-end agent. We propose a Data Synthesis Strategy Based on Atomic Capabilities to reinforce planning and report writing, combined with a progressive training path from agentic mid-training to SFT and RL. Enhanced by a Checklist-style Judger, this approach significantly improves robustness. Furthermore, to bridge the evaluation gap in the Chinese domain, we establish ADR-Bench for realistic deep research scenarios. Experimental results show that Step-DeepResearch (32B) scores 61.4% on Scale AI Research Rubrics. On ADR-Bench, it significantly outperforms comparable models and rivals SOTA closed-source models like OpenAI and Gemini DeepResearch. These findings prove that refined training enables medium-sized models to achieve expert-level capabilities at industry-leading cost-efficiency.

Bo Peng, Wujian Xu, Kun Wang et al.

Multi-window CT imaging captures complementary pathological information across anatomical structures of differing densities, yet existing deep learning methods fuse representations only at later stages, missing cross-density interactions. We propose a cross-window knowledge distillation framework in which student encoders learn latent clinical priors from a teacher trained on the most informative window. Evaluated retrospectively on three cohorts - COPD-CT-DF (n=719), RSNA PE (n=1,433), and an in-house CTEPD dataset (n=161) - distillation improved per-window AUC by 10.1-16.5 percentage points on COPD-CT-DF (0.75-0.81 to 0.90-0.94; all P<0.001), with ensemble AUC reaching 0.9960. Similar gains were observed on RSNA PE (0.80-0.83 to 0.90-0.92) and CTEPD (AUC 0.7481 vs. 0.6264). Cross-window distillation internalises pathological signatures invisible to supervised approaches, offering a generalisable solution for multi-window pulmonary CT analysis.

Lielin Jiang, Na Wang, Qingqing Dang et al.

Different from the Single Image Super-Resolution(SISR) task, the key for Video Super-Resolution(VSR) task is to make full use of complementary information across frames to reconstruct the high-resolution sequence. Since images from different frames with diverse motion and scene, accurately aligning multiple frames and effectively fusing different frames has always been the key research work of VSR tasks. To utilize rich complementary information of neighboring frames, in this paper, we propose a multi-stage VSR deep architecture, dubbed as PP-MSVSR, with local fusion module, auxiliary loss and re-align module to refine the enhanced result progressively. Specifically, in order to strengthen the fusion of features across frames in feature propagation, a local fusion module is designed in stage-1 to perform local feature fusion before feature propagation. Moreover, we introduce an auxiliary loss in stage-2 to make the features obtained by the propagation module reserve more correlated information connected to the HR space, and introduce a re-align module in stage-3 to make full use of the feature information of the previous stage. Extensive experiments substantiate that PP-MSVSR achieves a promising performance of Vid4 datasets, which achieves a PSNR of 28.13dB with only 1.45M parameters. And the PP-MSVSR-L exceeds all state of the art method on REDS4 datasets with considerable parameters. Code and models will be released in PaddleGAN\footnote{https://github.com/PaddlePaddle/PaddleGAN.}.

StepFun, Bin Wang, Bojun Wang et al.

Large language models (LLMs) face low hardware efficiency during decoding, especially for long-context reasoning tasks. This paper introduces Step-3, a 321B-parameter VLM with hardware-aware model-system co-design optimized for minimizing decoding costs. Step-3 innovates in two key dimensions: (1) A novel Multi-Matrix Factorization Attention (MFA) mechanism that significantly reduces both KV cache size and computation while maintaining high attention expressiveness, and (2) Attention-FFN Disaggregation (AFD), a distributed inference system that decouples attention and Feed-Forward Network (FFN) layers into specialized subsystems. This co-design achieves unprecedented cost efficiency: Step-3 significantly reduces theoretical decoding costs compared with models like DeepSeek-V3 and Qwen3 MoE 235B, with the gains widening at longer context. Step-3 achieves low cost while activating 38B parameters per token (more than DeepSeek-V3 and Qwen3 MoE 235B), demonstrating that hardware-aligned attention arithmetic intensity, MoE sparsity, and AFD are critical to cost-effectiveness. We perform a head-to-head comparison with DeepSeek-V3 in its favorable scenarios. Our implementation on Hopper GPUs achieves a decoding throughput of up to 4,039 tokens per second per GPU under 50ms TPOT SLA (4K context, FP8, no MTP). It is higher than DeepSeek-V3's 2,324 in the same setup and sets a new Pareto frontier for LLM decoding.

Na Wang, Lei Qi, Jintao Guo et al.

Domain generalization (DG) intends to train a model on multiple source domains to ensure that it can generalize well to an arbitrary unseen target domain. The acquisition of domain-invariant representations is pivotal for DG as they possess the ability to capture the inherent semantic information of the data, mitigate the influence of domain shift, and enhance the generalization capability of the model. Adopting multiple perspectives, such as the sample and the feature, proves to be effective. The sample perspective facilitates data augmentation through data manipulation techniques, whereas the feature perspective enables the extraction of meaningful generalization features. In this paper, we focus on improving the generalization ability of the model by compelling it to acquire domain-invariant representations from both the sample and feature perspectives by disentangling spurious correlations and enhancing potential correlations. 1) From the sample perspective, we develop a frequency restriction module, guiding the model to focus on the relevant correlations between object features and labels, thereby disentangling spurious correlations. 2) From the feature perspective, the simple Tail Interaction module implicitly enhances potential correlations among all samples from all source domains, facilitating the acquisition of domain-invariant representations across multiple domains for the model. The experimental results show that Convolutional Neural Networks (CNNs) or Multi-Layer Perceptrons (MLPs) with a strong baseline embedded with these two modules can achieve superior results, e.g., an average accuracy of 92.30% on Digits-DG.

Haolong Yan, Jia Wang, Xin Huang et al.

Recent advances in multimodal large language models unlock unprecedented opportunities for GUI automation. However, a fundamental challenge remains: how to efficiently acquire high-quality training data while maintaining annotation reliability? We introduce a self-evolving training pipeline powered by the Calibrated Step Reward System, which converts model-generated trajectories into reliable training signals through trajectory-level calibration, achieving >90% annotation accuracy with 10-100x lower cost. Leveraging this pipeline, we introduce Step-GUI, a family of models (4B/8B) that achieves state-of-the-art GUI performance (8B: 80.2% AndroidWorld, 48.5% OSWorld, 62.6% ScreenShot-Pro) while maintaining robust general capabilities. As GUI agent capabilities improve, practical deployment demands standardized interfaces across heterogeneous devices while protecting user privacy. To this end, we propose GUI-MCP, the first Model Context Protocol for GUI automation with hierarchical architecture that combines low-level atomic operations and high-level task delegation to local specialist models, enabling high-privacy execution where sensitive data stays on-device. Finally, to assess whether agents can handle authentic everyday usage, we introduce AndroidDaily, a benchmark grounded in real-world mobile usage patterns with 3146 static actions and 235 end-to-end tasks across high-frequency daily scenarios (8B: static 89.91%, end-to-end 52.50%). Our work advances the development of practical GUI agents and demonstrates strong potential for real-world deployment in everyday digital interactions.

Wenhui Lei, Qi Su, Ran Gu et al.

Deep neural networks usually require accurate and a large number of annotations to achieve outstanding performance in medical image segmentation. One-shot segmentation and weakly-supervised learning are promising research directions that lower labeling effort by learning a new class from only one annotated image and utilizing coarse labels instead, respectively. Previous works usually fail to leverage the anatomical structure and suffer from class imbalance and low contrast problems. Hence, we present an innovative framework for 3D medical image segmentation with one-shot and weakly-supervised settings. Firstly a propagation-reconstruction network is proposed to project scribbles from annotated volume to unlabeled 3D images based on the assumption that anatomical patterns in different human bodies are similar. Then a dual-level feature denoising module is designed to refine the scribbles based on anatomical- and pixel-level features. After expanding the scribbles to pseudo masks, we could train a segmentation model for the new class with the noisy label training strategy. Experiments on one abdomen and one head-and-neck CT dataset show the proposed method obtains significant improvement over the state-of-the-art methods and performs robustly even under severe class imbalance and low contrast.

Qi Duan, Guotai Wang, Rui Wang et al.

Clinical research on smart health has an increasing demand for intelligent and clinic-oriented medical image computing algorithms and platforms that support various applications. To this end, we have developed SenseCare research platform, which is designed to facilitate translational research on intelligent diagnosis and treatment planning in various clinical scenarios. To enable clinical research with Artificial Intelligence (AI), SenseCare provides a range of AI toolkits for different tasks, including image segmentation, registration, lesion and landmark detection from various image modalities ranging from radiology to pathology. In addition, SenseCare is clinic-oriented and supports a wide range of clinical applications such as diagnosis and surgical planning for lung cancer, pelvic tumor, coronary artery disease, etc. SenseCare provides several appealing functions and features such as advanced 3D visualization, concurrent and efficient web-based access, fast data synchronization and high data security, multi-center deployment, support for collaborative research, etc. In this report, we present an overview of SenseCare as an efficient platform providing comprehensive toolkits and high extensibility for intelligent image analysis and clinical research in different application scenarios. We also summarize the research outcome through the collaboration with multiple hospitals.

Xi Liu, Muhe Xie, Xidao Wen et al.

As mobile devices become more and more popular, mobile gaming has emerged as a promising market with billion-dollar revenues. A variety of mobile game platforms and services have been developed around the world. A critical challenge for these platforms and services is to understand the churn behavior in mobile games, which usually involves churn at micro level (between an app and a specific user) and macro level (between an app and all its users). Accurate micro-level churn prediction and macro-level churn ranking will benefit many stakeholders such as game developers, advertisers, and platform operators. In this paper, we present the first large-scale churn analysis for mobile games that supports both micro-level churn prediction and macro-level churn ranking. For micro-level churn prediction, in view of the common limitations of the state-of-the-art methods built upon traditional machine learning models, we devise a novel semi-supervised and inductive embedding model that jointly learns the prediction function and the embedding function for user-app relationships. We model these two functions by deep neural networks with a unique edge embedding technique that is able to capture both contextual information and relationship dynamics. We also design a novel attributed random walk technique that takes into consideration both topological adjacency and attribute similarities. To address macro-level churn ranking, we propose to construct a relationship graph with estimated micro-level churn probabilities as edge weights and adapt link analysis algorithms on the graph. We devise a simple algorithm SimSum and adapt two more advanced algorithms PageRank and HITS. The performance of our solutions for the two-level churn analysis problems is evaluated on real-world data collected from the Samsung Game Launcher platform.

Xin Yang, Na Wang, Yi Wang et al.

Segmenting left atrium in MR volume holds great potentials in promoting the treatment of atrial fibrillation. However, the varying anatomies, artifacts and low contrasts among tissues hinder the advance of both manual and automated solutions. In this paper, we propose a fully-automated framework to segment left atrium in gadolinium-enhanced MR volumes. The region of left atrium is firstly automatically localized by a detection module. Our framework then originates with a customized 3D deep neural network to fully explore the spatial dependency in the region for segmentation. To alleviate the risk of low training efficiency and potential overfitting, we enhance our deep network with the transfer learning and deep supervision strategy. Main contribution of our network design lies in the composite loss function to combat the boundary ambiguity and hard examples. We firstly adopt the Overlap loss to encourage network reduce the overlap between the foreground and background and thus sharpen the predictions on boundary. We then propose a novel Focal Positive loss to guide the learning of voxel-specific threshold and emphasize the foreground to improve classification sensitivity. Further improvement is obtained with an recursive training scheme. With ablation studies, all the introduced modules prove to be effective. The proposed framework achieves an average Dice of 92.24 in segmenting left atrium with pulmonary veins on 20 testing volumes.

Xi Liu, Muhe Xie, Xidao Wen et al.

Mobile gaming has emerged as a promising market with billion-dollar revenues. A variety of mobile game platforms and services have been developed around the world. One critical challenge for these platforms and services is to understand user churn behavior in mobile games. Accurate churn prediction will benefit many stakeholders such as game developers, advertisers, and platform operators. In this paper, we present the first large-scale churn prediction solution for mobile games. In view of the common limitations of the state-of-the-art methods built upon traditional machine learning models, we devise a novel semi-supervised and inductive embedding model that jointly learns the prediction function and the embedding function for user-app relationships. We model these two functions by deep neural networks with a unique edge embedding technique that is able to capture both contextual information and relationship dynamics. We also design a novel attributed random walk technique that takes into consideration both topological adjacency and attribute similarities. To evaluate the performance of our solution, we collect real-world data from the Samsung Game Launcher platform that includes tens of thousands of games and hundreds of millions of user-app interactions. The experimental results with this data demonstrate the superiority of our proposed model against existing state-of-the-art methods.