Zheng Gong

Junyi Li, Tianyi Tang, Zheng Gong et al. · pku

Nowadays, pretrained language models (PLMs) have dominated the majority of NLP tasks. While, little research has been conducted on systematically evaluating the language abilities of PLMs. In this paper, we present a large-scale empirical study on general language ability evaluation of PLMs (ElitePLM). In our study, we design four evaluation dimensions, i.e. memory, comprehension, reasoning, and composition, to measure ten widely-used PLMs within five categories. Our empirical results demonstrate that: (1) PLMs with varying training objectives and strategies are good at different ability tests; (2) fine-tuning PLMs in downstream tasks is usually sensitive to the data size and distribution; (3) PLMs have excellent transferability between similar tasks. Moreover, the prediction results of PLMs in our experiments are released as an open resource for more deep and detailed analysis on the language abilities of PLMs. This paper can guide the future work to select, apply, and design PLMs for specific tasks. We have made all the details of experiments publicly available at https://github.com/RUCAIBox/ElitePLM.

Wayne Xin Zhao, Kun Zhou, Zheng Gong et al.

This paper aims to advance the mathematical intelligence of machines by presenting the first Chinese mathematical pre-trained language model~(PLM) for effectively understanding and representing mathematical problems. Unlike other standard NLP tasks, mathematical texts are difficult to understand, since they involve mathematical terminology, symbols and formulas in the problem statement. Typically, it requires complex mathematical logic and background knowledge for solving mathematical problems. Considering the complex nature of mathematical texts, we design a novel curriculum pre-training approach for improving the learning of mathematical PLMs, consisting of both basic and advanced courses. Specially, we first perform token-level pre-training based on a position-biased masking strategy, and then design logic-based pre-training tasks that aim to recover the shuffled sentences and formulas, respectively. Finally, we introduce a more difficult pre-training task that enforces the PLM to detect and correct the errors in its generated solutions. We conduct extensive experiments on offline evaluation (including nine math-related tasks) and online $A/B$ test. Experimental results demonstrate the effectiveness of our approach compared with a number of competitive baselines. Our code is available at: \textcolor{blue}{\url{https://github.com/RUCAIBox/JiuZhang}}.

Ruihang Miao, Weizhou Liu, Mingrui Chen et al.

3D Semantic Scene Completion (SSC) can provide dense geometric and semantic scene representations, which can be applied in the field of autonomous driving and robotic systems. It is challenging to estimate the complete geometry and semantics of a scene solely from visual images, and accurate depth information is crucial for restoring 3D geometry. In this paper, we propose the first stereo SSC method named OccDepth, which fully exploits implicit depth information from stereo images (or RGBD images) to help the recovery of 3D geometric structures. The Stereo Soft Feature Assignment (Stereo-SFA) module is proposed to better fuse 3D depth-aware features by implicitly learning the correlation between stereo images. In particular, when the input are RGBD image, a virtual stereo images can be generated through original RGB image and depth map. Besides, the Occupancy Aware Depth (OAD) module is used to obtain geometry-aware 3D features by knowledge distillation using pre-trained depth models. In addition, a reformed TartanAir benchmark, named SemanticTartanAir, is provided in this paper for further testing our OccDepth method on SSC task. Compared with the state-of-the-art RGB-inferred SSC method, extensive experiments on SemanticKITTI show that our OccDepth method achieves superior performance with improving +4.82% mIoU, of which +2.49% mIoU comes from stereo images and +2.33% mIoU comes from our proposed depth-aware method. Our code and trained models are available at https://github.com/megvii-research/OccDepth.

Weihao Gao, Zhuo Deng, Zhiyuan Niu et al.

Large multimodal language models (LMMs) have achieved significant success in general domains. However, due to the significant differences between medical images and text and general web content, the performance of LMMs in medical scenarios is limited. In ophthalmology, clinical diagnosis relies on multiple modalities of medical images, but unfortunately, multimodal ophthalmic large language models have not been explored to date. In this paper, we study and construct an ophthalmic large multimodal model. Firstly, we use fundus images as an entry point to build a disease assessment and diagnosis pipeline to achieve common ophthalmic disease diagnosis and lesion segmentation. Then, we establish a new ophthalmic multimodal instruction-following and dialogue fine-tuning dataset based on disease-related knowledge data and publicly available real-world medical dialogue. We introduce visual ability into the large language model to complete the ophthalmic large language and vision assistant (OphGLM). Our experimental results demonstrate that the OphGLM model performs exceptionally well, and it has the potential to revolutionize clinical applications in ophthalmology. The dataset, code, and models will be made publicly available at https://github.com/ML-AILab/OphGLM.

Wayne Xin Zhao, Kun Zhou, Beichen Zhang et al.

Although pre-trained language models~(PLMs) have recently advanced the research progress in mathematical reasoning, they are not specially designed as a capable multi-task solver, suffering from high cost for multi-task deployment (\eg a model copy for a task) and inferior performance on complex mathematical problems in practical applications. To address these issues, in this paper, we propose \textbf{JiuZhang~2.0}, a unified Chinese PLM specially for multi-task mathematical problem solving. Our idea is to maintain a moderate-sized model and employ the \emph{cross-task knowledge sharing} to improve the model capacity in a multi-task setting. Specially, we construct a Mixture-of-Experts~(MoE) architecture for modeling mathematical text, so as to capture the common mathematical knowledge across tasks. For optimizing the MoE architecture, we design \emph{multi-task continual pre-training} and \emph{multi-task fine-tuning} strategies for multi-task adaptation. These training strategies can effectively decompose the knowledge from the task data and establish the cross-task sharing via expert networks. In order to further improve the general capacity of solving different complex tasks, we leverage large language models~(LLMs) as complementary models to iteratively refine the generated solution by our PLM, via in-context learning. Extensive experiments have demonstrated the effectiveness of our model.

Qi Liu, Zheng Gong, Zhenya Huang et al.

Machine learning algorithms have become ubiquitous in a number of applications (e.g. image classification). However, due to the insufficient measurement of traditional metrics (e.g. the coarse-grained Accuracy of each classifier), substantial gaps are usually observed between the real-world performance of these algorithms and their scores in standardized evaluations. In this paper, inspired by the psychometric theories from human measurement, we propose a task-agnostic evaluation framework Camilla, where a multi-dimensional diagnostic metric Ability is defined for collaboratively measuring the multifaceted strength of each machine learning algorithm. Specifically, given the response logs from different algorithms to data samples, we leverage cognitive diagnosis assumptions and neural networks to learn the complex interactions among algorithms, samples and the skills (explicitly or implicitly pre-defined) of each sample. In this way, both the abilities of each algorithm on multiple skills and some of the sample factors (e.g. sample difficulty) can be simultaneously quantified. We conduct extensive experiments with hundreds of machine learning algorithms on four public datasets, and our experimental results demonstrate that Camilla not only can capture the pros and cons of each algorithm more precisely, but also outperforms state-of-the-art baselines on the metric reliability, rank consistency and rank stability.

Milan Ganai, Zheng Gong, Chenning Yu et al.

Ensuring safety is important for the practical deployment of reinforcement learning (RL). Various challenges must be addressed, such as handling stochasticity in the environments, providing rigorous guarantees of persistent state-wise safety satisfaction, and avoiding overly conservative behaviors that sacrifice performance. We propose a new framework, Reachability Estimation for Safe Policy Optimization (RESPO), for safety-constrained RL in general stochastic settings. In the feasible set where there exist violation-free policies, we optimize for rewards while maintaining persistent safety. Outside this feasible set, our optimization produces the safest behavior by guaranteeing entrance into the feasible set whenever possible with the least cumulative discounted violations. We introduce a class of algorithms using our novel reachability estimation function to optimize in our proposed framework and in similar frameworks such as those concurrently handling multiple hard and soft constraints. We theoretically establish that our algorithms almost surely converge to locally optimal policies of our safe optimization framework. We evaluate the proposed methods on a diverse suite of safe RL environments from Safety Gym, PyBullet, and MuJoCo, and show the benefits in improving both reward performance and safety compared with state-of-the-art baselines.

Yuting Ning, Zhenya Huang, Xin Lin et al.

Understanding mathematical questions effectively is a crucial task, which can benefit many applications, such as difficulty estimation. Researchers have drawn much attention to designing pre-training models for question representations due to the scarcity of human annotations (e.g., labeling difficulty). However, unlike general free-format texts (e.g., user comments), mathematical questions are generally designed with explicit purposes and mathematical logic, and usually consist of more complex content, such as formulas, and related mathematical knowledge (e.g., Function). Therefore, the problem of holistically representing mathematical questions remains underexplored. To this end, in this paper, we propose a novel contrastive pre-training approach for mathematical question representations, namely QuesCo, which attempts to bring questions with more similar purposes closer. Specifically, we first design two-level question augmentations, including content-level and structure-level, which generate literally diverse question pairs with similar purposes. Then, to fully exploit hierarchical information of knowledge concepts, we propose a knowledge hierarchy-aware rank strategy (KHAR), which ranks the similarities between questions in a fine-grained manner. Next, we adopt a ranking contrastive learning task to optimize our model based on the augmented and ranked questions. We conduct extensive experiments on two real-world mathematical datasets. The experimental results demonstrate the effectiveness of our model.

Weihao Gao, Zheng Gong, Zhuo Deng et al.

Tabular data is the most common type of data in real-life scenarios. In this study, we propose the TabKANet model for tabular data modeling, which targets the bottlenecks in learning from numerical content. We constructed a Kolmogorov-Arnold Network (KAN) based Numerical Embedding Module and unified numerical and categorical features encoding within a Transformer architecture. TabKANet has demonstrated stable and significantly superior performance compared to Neural Networks (NNs) across multiple public datasets in binary classification, multi-class classification, and regression tasks. Its performance is comparable to or surpasses that of Gradient Boosted Decision Tree models (GBDTs). Our code is publicly available on GitHub: https://github.com/AI-thpremed/TabKANet.

Boyang Li, Zheng Gong, Sylvia Herbert

In this article, we consider the infinite-horizon reach-avoid (RA) and stabilize-avoid (SA) zero-sum game problems for general nonlinear continuous-time systems, where the goal is to find the set of states that can be controlled to reach or stabilize to a target set, without violating constraints even under the worst-case disturbance. Based on the Hamilton-Jacobi reachability method, we address the RA problem by designing a new Lipschitz continuous RA value function, whose zero sublevel set exactly characterizes the RA set. We establish that the associated Bellman backup operator is contractive and that the RA value function is the unique viscosity solution of a Hamilton-Jacobi variational inequality. Finally, we develop a two-step framework for the SA problem by integrating our RA strategies with a recently proposed Robust Control Lyapunov-Value Function, thereby ensuring both target reachability and long-term stability. We numerically verify our RA and SA frameworks on a 3D Dubins car system to demonstrate the efficacy of the proposed approach.

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.

Feng Gao, Zheng Gong, Wenli Liu et al.

While Mamba models offer efficient sequence modeling, vanilla versions struggle with temporal correlations and boundary details in Arctic sea ice concentration (SIC) prediction. To address these limitations, we propose Frequency-enhanced Hilbert scanning Mamba Framework (FH-Mamba) for short-term Arctic SIC prediction. Specifically, we introduce a 3D Hilbert scan mechanism that traverses the 3D spatiotemporal grid along a locality-preserving path, ensuring that adjacent indices in the flattened sequence correspond to neighboring voxels in both spatial and temporal dimensions. Additionally, we incorporate wavelet transform to amplify high-frequency details and we also design a Hybrid Shuffle Attention module to adaptively aggregate sequence and frequency features. Experiments conducted on the OSI-450a1 and AMSR2 datasets demonstrate that our FH-Mamba achieves superior prediction performance compared with state-of-the-art baselines. The results confirm the effectiveness of Hilbert scanning and frequency-aware attention in improving both temporal consistency and edge reconstruction for Arctic SIC forecasting. Our codes are publicly available at https://github.com/oucailab/FH-Mamba.

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.

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.

Yan Xiang, Yu-Hang Tang, Zheng Gong et al.

We introduce an explorative active learning (AL) algorithm based on Gaussian process regression and marginalized graph kernel (GPR-MGK) to explore chemical space with minimum cost. Using high-throughput molecular dynamics simulation to generate data and graph neural network (GNN) to predict, we constructed an active learning molecular simulation framework for thermodynamic property prediction. In specific, targeting 251,728 alkane molecules consisting of 4 to 19 carbon atoms and their liquid physical properties: densities, heat capacities, and vaporization enthalpies, we use the AL algorithm to select the most informative molecules to represent the chemical space. Validation of computational and experimental test sets shows that only 313 (0.124\% of the total) molecules were sufficient to train an accurate GNN model with $\rm R^2 > 0.99$ for computational test sets and $\rm R^2 > 0.94$ for experimental test sets. We highlight two advantages of the presented AL algorithm: compatibility with high-throughput data generation and reliable uncertainty quantification.

Yang Ji, Ying Sun, Yuting Zhang et al.

Neural networks have achieved remarkable success across various fields. However, the lack of interpretability limits their practical use, particularly in critical decision-making scenarios. Post-hoc interpretability, which provides explanations for pre-trained models, is often at risk of robustness and fidelity. This has inspired a rising interest in self-interpretable neural networks, which inherently reveal the prediction rationale through the model structures. Although there exist surveys on post-hoc interpretability, a comprehensive and systematic survey of self-interpretable neural networks is still missing. To address this gap, we first collect and review existing works on self-interpretable neural networks and provide a structured summary of their methodologies from five key perspectives: attribution-based, function-based, concept-based, prototype-based, and rule-based self-interpretation. We also present concrete, visualized examples of model explanations and discuss their applicability across diverse scenarios, including image, text, graph data, and deep reinforcement learning. Additionally, we summarize existing evaluation metrics for self-interpretability and identify open challenges in this field, offering insights for future research. To support ongoing developments, we present a publicly accessible resource to track advancements in this domain: https://github.com/yangji721/Awesome-Self-Interpretable-Neural-Network.

Sha Yuan, Shuai Zhao, Jiahong Leng et al.

Compared with the domain-specific model, the vision-language pre-training models (VLPMs) have shown superior performance on downstream tasks with fast fine-tuning process. For example, ERNIE-ViL, Oscar and UNIMO trained VLPMs with a uniform transformers stack architecture and large amounts of image-text paired data, achieving remarkable results on downstream tasks such as image-text reference(IR and TR), vision question answering (VQA) and image captioning (IC) etc. During the training phase, VLPMs are always fed with a combination of multiple public datasets to meet the demand of large-scare training data. However, due to the unevenness of data distribution including size, task type and quality, using the mixture of multiple datasets for model training can be problematic. In this work, we introduce a large-scale multi-modal corpora named WuDaoMM, totally containing more than 650M image-text pairs. Specifically, about 600 million pairs of data are collected from multiple webpages in which image and caption present weak correlation, and the other 50 million strong-related image-text pairs are collected from some high-quality graphic websites. We also release a base version of WuDaoMM with 5 million strong-correlated image-text pairs, which is sufficient to support the common cross-modal model pre-training. Besides, we trained both an understanding and a generation vision-language (VL) model to test the dataset effectiveness. The results show that WuDaoMM can be applied as an efficient dataset for VLPMs, especially for the model in text-to-image generation task. The data is released at https://data.wudaoai.cn

Zheng Gong, Ziyi Jiang, Weihao Gao et al.

The mRNA optimization is critical for therapeutic and biotechnological applications, since sequence features directly govern protein expression levels and efficacy. However, current methods face significant challenges in simultaneously achieving three key objectives: (1) fidelity (preventing unintended amino acid changes), (2) computational efficiency (speed and scalability), and (3) the scope of optimization variables considered (multi-objective capability). Furthermore, existing methods often fall short of comprehensively incorporating the factors related to the mRNA lifecycle and translation process, including intrinsic mRNA sequence properties, secondary structure, translation elongation kinetics, and tRNA availability. To address these limitations, we introduce \textbf{RNop}, a novel deep learning-based method for mRNA optimization. We collect a large-scale dataset containing over 3 million sequences and design four specialized loss functions, the GPLoss, CAILoss, tAILoss, and MFELoss, which simultaneously enable explicit control over sequence fidelity while optimizing species-specific codon adaptation, tRNA availability, and desirable mRNA secondary structure features. Then, we demonstrate RNop's effectiveness through extensive in silico and in vivo experiments. RNop ensures high sequence fidelity, achieves significant computational throughput up to 47.32 sequences/s, and yields optimized mRNA sequences resulting in a significant increase in protein expression for functional proteins compared to controls. RNop surpasses current methodologies in both quantitative metrics and experimental validation, enlightening a new dawn for efficient and effective mRNA design. Code and models will be available at https://github.com/HudenJear/RPLoss.

Weihao Gao, Zhuo Deng, Zheng Gong et al.

Early detection of colorectal cancer hinges on real-time, accurate polyp identification and resection. Yet current high-precision segmentation models rely on GPUs, making them impractical to deploy in primary hospitals, mobile endoscopy units, or capsule robots. To bridge this gap, we present the UltraSeg family, operating in an extreme-compression regime (<0.3 M parameters). UltraSeg-108K (0.108 M parameters) is optimized for single-center data, while UltraSeg-130K (0.13 M parameters) generalizes to multi-center, multi-modal images. By jointly optimizing encoder-decoder widths, incorporating constrained dilated convolutions to enlarge receptive fields, and integrating a cross-layer lightweight fusion module, the models achieve 90 FPS on a single CPU core without sacrificing accuracy. Evaluated on seven public datasets, UltraSeg retains >94% of the Dice score of a 31 M-parameter U-Net while utilizing only 0.4% of its parameters, establishing a strong, clinically viable baseline for the extreme-compression domain and offering an immediately deployable solution for resource-constrained settings. This work provides not only a CPU-native solution for colonoscopy but also a reproducible blueprint for broader minimally invasive surgical vision applications. Source code is publicly available to ensure reproducibility and facilitate future benchmarking.

Zhipeng Chen, Kun Zhou, Beichen Zhang et al.

Although large language models (LLMs) have achieved excellent performance in a variety of evaluation benchmarks, they still struggle in complex reasoning tasks which require specific knowledge and multi-hop reasoning. To improve the reasoning abilities, we propose ChatCoT, a tool-augmented chain-of-thought reasoning framework for chat-based LLMs (e.g., ChatGPT). In ChatCoT, we model the chain-of-thought (CoT) reasoning as multi-turn conversations, to utilize tools in a more natural way through chatting. At each turn, LLMs can either interact with tools or perform the reasoning. Our approach can effectively leverage the multi-turn conversation ability of chat-based LLMs, and integrate the thought chain following and tools manipulation in a unified way. Specially, we initialize the early turns of the conversation by the knowledge about tools, tasks, and reasoning format, and propose an iterative tool-augmented reasoning step to perform step-by-step tool-augmented reasoning. The experiment results on two complex reasoning datasets (MATH and HotpotQA) have shown the effectiveness of ChatCoT on complex reasoning tasks, achieving a 7.9% relative improvement over the state-of-the-art baseline. Our code and data are available at: \url{https://github.com/RUCAIBOX/ChatCoT}.

Zhuo Deng, Yuanhao Cai, Lu Chen et al.

Ophthalmologists have used fundus images to screen and diagnose eye diseases. However, different equipments and ophthalmologists pose large variations to the quality of fundus images. Low-quality (LQ) degraded fundus images easily lead to uncertainty in clinical screening and generally increase the risk of misdiagnosis. Thus, real fundus image restoration is worth studying. Unfortunately, real clinical benchmark has not been explored for this task so far. In this paper, we investigate the real clinical fundus image restoration problem. Firstly, We establish a clinical dataset, Real Fundus (RF), including 120 low- and high-quality (HQ) image pairs. Then we propose a novel Transformer-based Generative Adversarial Network (RFormer) to restore the real degradation of clinical fundus images. The key component in our network is the Window-based Self-Attention Block (WSAB) which captures non-local self-similarity and long-range dependencies. To produce more visually pleasant results, a Transformer-based discriminator is introduced. Extensive experiments on our clinical benchmark show that the proposed RFormer significantly outperforms the state-of-the-art (SOTA) methods. In addition, experiments of downstream tasks such as vessel segmentation and optic disc/cup detection demonstrate that our proposed RFormer benefits clinical fundus image analysis and applications. The dataset, code, and models are publicly available at https://github.com/dengzhuo-AI/Real-Fundus

Wen Yang, Zheng Gong, Baifu Huang et al.

Lidar point cloud distortion from moving object is an important problem in autonomous driving, and recently becomes even more demanding with the emerging of newer lidars, which feature back-and-forth scanning patterns. Accurately estimating moving object velocity would not only provide a tracking capability but also correct the point cloud distortion with more accurate description of the moving object. Since lidar measures the time-of-flight distance but with a sparse angular resolution, the measurement is precise in the radial measurement but lacks angularly. Camera on the other hand provides a dense angular resolution. In this paper, Gaussian-based lidar and camera fusion is proposed to estimate the full velocity and correct the lidar distortion. A probabilistic Kalman-filter framework is provided to track the moving objects, estimate their velocities and simultaneously correct the point clouds distortions. The framework is evaluated on real road data and the fusion method outperforms the traditional ICP-based and point-cloud only method. The complete working framework is open-sourced (https://github.com/ISEE-Technology/lidar-with-velocity) to accelerate the adoption of the emerging lidars.

Shunzhi Yang, Zheng Gong, Kai Ye et al.

With the development of Internet of Things (IoT), data is increasingly appearing on the edge of the network. Processing tasks on the edge of the network can effectively solve the problems of personal privacy leaks and server overload. As a result, it has attracted a great deal of attention and made substantial progress. This progress includes efficient convolutional neural network (CNN) models such as MobileNet and ShuffleNet. However, all of these networks appear as a common network model and they usually need to identify multiple targets when applied. So the size of the input is very large. In some specific cases, only the target needs to be classified. Therefore, a small input network can be designed to reduce computation. In addition, other efficient neural network models are primarily designed for mobile phones. Mobile phones have faster memory access, which allows them to use group convolution. In particular, this paper finds that the recently widely used group convolution is not suitable for devices with very slow memory access. Therefore, the EdgeCNN of this paper is designed for edge computing devices with low memory access speed and low computing resources. EdgeCNN has been run successfully on the Raspberry Pi 3B+ at a speed of 1.37 frames per second. The accuracy of facial expression classification for the FER-2013 and RAF-DB datasets outperforms other proposed networks that are compatible with the Raspberry Pi 3B+. The implementation of EdgeCNN is available at https://github.com/yangshunzhi1994/EdgeCNN

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.

Ailin Huang, Bingxin Li, Bruce Wang et al.

Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a dual-codebook audio tokenizer for linguistic and semantic feature extraction, a 130-billion-parameter backbone LLM and a neural vocoder for high-fidelity speech synthesis. Our post-training approach employs interleaved token-output of text and audio to enhance semantic coherence and combines Direct Preference Optimization (DPO) with model merge to improve performance. Evaluations on the StepEval-Audio-360 benchmark demonstrate that Step-Audio-AQAA excels especially in speech control, outperforming the state-of-art LALMs in key areas. This work contributes a promising solution for end-to-end LALMs and highlights the critical role of token-based vocoder in enhancing overall performance for AQAA tasks.

Weihao Gao, Zhuo Deng, Zheng Gong et al.

Insulin resistance (IR) is a key precursor to diabetes and a significant risk factor for cardiovascular disease. Traditional IR assessment methods require multiple blood tests. We developed a simple AI model using only fasting blood glucose to predict IR in non-diabetic populations. Data from the NHANES (1999-2020) and CHARLS (2015) studies were used for model training and validation. Input features included age, gender, height, weight, blood pressure, waist circumference, and fasting blood glucose. The CatBoost algorithm achieved AUC values of 0.8596 (HOMA-IR) and 0.7777 (TyG index) in NHANES, with an external AUC of 0.7442 for TyG. For METS-IR prediction, the model achieved AUC values of 0.9731 (internal) and 0.9591 (external), with RMSE values of 3.2643 (internal) and 3.057 (external). SHAP analysis highlighted waist circumference as a key predictor of IR. This AI model offers a minimally invasive and effective tool for IR prediction, supporting early diabetes and cardiovascular disease prevention.

Zheng Gong, Zhuo Deng, Run Gan et al.

The retinal fundus images are utilized extensively in the diagnosis, and their quality can directly affect the diagnosis results. However, due to the insufficient dataset and algorithm application, current fundus image quality assessment (FIQA) methods are not powerful enough to meet ophthalmologists` demands. In this paper, we address the limitations of datasets and algorithms in FIQA. First, we establish a new FIQA dataset, Fundus Quality Score(FQS), which includes 2246 fundus images with two labels: a continuous Mean Opinion Score varying from 0 to 100 and a three-level quality label. Then, we propose a FIQA Transformer-based Hypernetwork (FTHNet) to solve these tasks with regression results rather than classification results in conventional FIQA works. The FTHNet is optimized for the FIQA tasks with extensive experiments. Results on our FQS dataset show that the FTHNet can give quality scores for fundus images with PLCC of 0.9423 and SRCC of 0.9488, significantly outperforming other methods with fewer parameters and less computation complexity.We successfully build a dataset and model addressing the problems of current FIQA methods. Furthermore, the model deployment experiments demonstrate its potential in automatic medical image quality control. All experiments are carried out with 10-fold cross-validation to ensure the significance of the results.

Zihao Wang, Yunjie Li, Lingmin Zan et al.

The Extended Long Short-Term Memory (xLSTM) network has attracted widespread research interest due to its enhanced capability to model complex temporal dependencies in diverse time series applications. Despite its success, there is still potential to further improve its representational capacity and forecasting performance, particularly on challenging real-world datasets with unknown, intricate, and hierarchical dynamics. In this work, we propose a stochastic xLSTM, termed StoxLSTM, that improves the original architecture into a state space modeling framework by incorporating stochastic latent variables within xLSTM. StoxLSTM models the latent dynamic evolution through specially designed recurrent blocks, enabling it to effectively capture the underlying temporal patterns and dependencies. Extensive experiments on publicly available benchmark datasets from multiple research communities demonstrate that StoxLSTM consistently outperforms state-of-the-art baselines with better robustness and stronger generalization ability.

Zheng Gong, Zhuo Deng, Weihao Gao et al.

Cataract is one of the most common blinding eye diseases and can be treated by surgery. However, because cataract patients may also suffer from other blinding eye diseases, ophthalmologists must diagnose them before surgery. The cloudy lens of cataract patients forms a hazy degeneration in the fundus images, making it challenging to observe the patient's fundus vessels, which brings difficulties to the diagnosis process. To address this issue, this paper establishes a new cataract image restoration method named Catintell. It contains a cataract image synthesizing model, Catintell-Syn, and a restoration model, Catintell-Res. Catintell-Syn uses GAN architecture with fully unsupervised data to generate paired cataract-like images with realistic style and texture rather than the conventional Gaussian degradation algorithm. Meanwhile, Catintell-Res is an image restoration network that can improve the quality of real cataract fundus images using the knowledge learned from synthetic cataract images. Extensive experiments show that Catintell-Res outperforms other cataract image restoration methods in PSNR with 39.03 and SSIM with 0.9476. Furthermore, the universal restoration ability that Catintell-Res gained from unpaired cataract images can process cataract images from various datasets. We hope the models can help ophthalmologists identify other blinding eye diseases of cataract patients and inspire more medical image restoration methods in the future.

Waqas Ali, Peilin Liu, Rendong Ying et al.

Most real-time autonomous robot applications require a robot to traverse through a dynamic space for a long time. In some cases, a robot needs to work in the same environment. Such applications give rise to the problem of a life-long SLAM system. Life-long SLAM presents two main challenges i.e. the tracking should not fail in a dynamic environment and the need for a robust and efficient mapping strategy. The system should update maps with new information; while also keeping track of older observations. But, mapping for a long time can require higher computational requirements. In this paper, we propose a solution to the problem of life-long SLAM. We represent the global map as a set of rasterized images of local maps along with a map management system responsible for updating local maps and keeping track of older values. We also present an efficient approach of using the bag of visual words method for loop closure detection and relocalization. We evaluate the performance of our system on the KITTI dataset and an indoor dataset. Our loop closure system reported recall and precision of above 90 percent. The computational cost of our system is much lower as compared to state-of-the-art methods. Our method reports lower computational requirements even for long-term operation.

Waqas Ali, Peilin Liu, Rendong Ying et al.

An accurate and computationally efficient SLAM algorithm is vital for modern autonomous vehicles. To make a lightweight the algorithm, most SLAM systems rely on feature detection from images for vision SLAM or point cloud for laser-based methods. Feature detection through a 3D point cloud becomes a computationally challenging task. In this paper, we propose a feature detection method by projecting a 3D point cloud to form an image and apply the vision-based feature detection technique. The proposed method gives repeatable and stable features in a variety of environments. Based on such features, we build a 6-DOF SLAM system consisting of tracking, mapping, and loop closure threads. For loop detection, we employ a 2-step approach i.e. nearest key-frames detection and loop candidate verification by matching features extracted from rasterized LIDAR images. Furthermore, we utilize a key-frame structure to achieve a lightweight SLAM system. The proposed system is evaluated with implementation on the KITTI dataset and the University of Michigan Ford Campus dataset. Through experimental results, we show that the algorithm presented in this paper can substantially reduce the computational cost of feature detection from the point cloud and the whole SLAM system while giving accurate results.

Yuqi Huo, Manli Zhang, Guangzhen Liu et al.

Multi-modal pre-training models have been intensively explored to bridge vision and language in recent years. However, most of them explicitly model the cross-modal interaction between image-text pairs, by assuming that there exists strong semantic correlation between the text and image modalities. Since this strong assumption is often invalid in real-world scenarios, we choose to implicitly model the cross-modal correlation for large-scale multi-modal pre-training, which is the focus of the Chinese project `WenLan' led by our team. Specifically, with the weak correlation assumption over image-text pairs, we propose a two-tower pre-training model called BriVL within the cross-modal contrastive learning framework. Unlike OpenAI CLIP that adopts a simple contrastive learning method, we devise a more advanced algorithm by adapting the latest method MoCo into the cross-modal scenario. By building a large queue-based dictionary, our BriVL can incorporate more negative samples in limited GPU resources. We further construct a large Chinese multi-source image-text dataset called RUC-CAS-WenLan for pre-training our BriVL model. Extensive experiments demonstrate that the pre-trained BriVL model outperforms both UNITER and OpenAI CLIP on various downstream tasks.