Ziwen Wang

Haoyang Ge, Jian Ma, Ziwen Wang et al.

High-quality dynamic human pose annotation equips AI with precise motion kinematics to enable human behavior mastery, yet remains labor-intensive and time-consuming. Current annotation tools either lack temporal correction propagation or fail in multi-person scenarios, necessitating excessive manual intervention. In this paper, we introduce IMPose, an interactive tool for multi-person dynamic pose annotation. It features a dual-level tracking mechanism that propagates one-frame multi-person pose corrections from annotators across entire videos. The keypoint-level ensures corrections temporal propagation via sequential modeling, while the instance-level employs keypoint-aware embedding with relative positional encoding to maintain multi-person cross-frame consistency. To further improve robustness, IMPose maintains historical pose and instance cues in a trajectory bank, which enhances long-range temporal association and stabilizes annotation in challenging cases such as occlusion and motion blur. By converting sparse human corrections into dense and coherent pose trajectories, our framework significantly reduces repeated manual refinement across frames. Extensive experiments show that IMPose consistently achieves a strong accuracy efficiency trade off under different interaction budgets, demonstrating particular advantages in low click annotation settings. IMPose achieves high precision annotation with high efficiency, requiring only 27 clicks per 1,050 frame video on 3DPW and 3 clicks per tracklet per 84-frame on PoseTrack21. We further expand PoseTrack21 with 188K pose instances (3.55M keypoints) at a minimal cost of 10 annotators in 10 hours. The annotation tool, codes, and extended dataset will be open-sourced.

Deli Huang, Cunguang Wang, Hongyin Tang et al.

Long-context language models now advertise context windows up to millions of tokens, yet evaluations typically report a single length or a narrow task family, masking two failure modes: performance can collapse as length grows, and strong retrieval need not transfer to downstream use. We present ATLAS, a benchmarking framework that redefines long-context evaluation as length-dependent capability profiling. ATLAS contributes three methodological principles:(i) a layered taxonomy separating foundational operations from application workloads so failures can be attributed, (ii) length-aware AUC scoring that integrates score-length curves over a fixed 8K-1M grid, replacing single-point metrics with full degradation profiles, and (iii) ATLAScore, a harmonic-mean aggregate over taxonomy categories that penalizes imbalanced profiles, with end-to-end uncertainty propagation from subset scores through the nonlinear final aggregate. We instantiate the framework across eight capability dimensions with nine auditable components and 6,438 instances, and evaluate 26 models. Gemini-3.1-Pro-Preview leads at 128K, Claude-Opus-4.6 leads at 1M. Rankings reshuffle substantially between ATLASscore@8K-128K and ATLASscore@8K-1M: 7 models move by at least two ranks, and the two taxonomy layers share only 61% of cross-model variance, with individual rank gaps up to 12 positions. These results support reporting long-context quality by capability and length, not by a single headline score.

Kaining Ying, Hengrui Hu, Siyu Ren et al.

Interactive world models are advancing rapidly, yet existing benchmarks cover only part of the required competencies, leaving no unified standard for systematic evaluation. To fill this gap, we introduce WBench, a comprehensive multi-turn benchmark for interactive world model evaluation along five dimensions, namely video quality, setting adherence, interaction adherence, consistency, and physics compliance. WBench contains 289 test cases and 1,058 interaction turns, where each case specifies a world setting and a multi-turn interaction sequence, covering diverse scenes, styles, subjects, and both first- and third-person perspectives, together with four interaction types, including navigation, subject action, event editing, and perspective switching. For navigation, WBench unifies text, 6-DoF pose, and discrete-action control, enabling evaluation of models with different native input interfaces. Evaluation uses 22 automatic sub-metrics that combine specialist vision models with large multimodal models, and all metrics are validated against human judgments. Across 20 state-of-the-art models, we find that no single model performs strongly across all dimensions. We provide detailed diagnostic insights into the characteristic strengths, weaknesses, and open challenges of each model. Code and data are available at https://github.com/meituan-longcat/WBench.

Meituan LongCat Team, Anchun Gui, Bei Li et al.

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

Meituan LongCat Team, Bin Xiao, Chao Wang et al.

The prevailing Next-Token Prediction (NTP) paradigm has driven the success of large language models through discrete autoregressive modeling. However, contemporary multimodal systems remain language-centric, often treating non-linguistic modalities as external attachments, leading to fragmented architectures and suboptimal integration. To transcend this limitation, we introduce Discrete Native Autoregressive (DiNA), a unified framework that represents multimodal information within a shared discrete space, enabling a consistent and principled autoregressive modeling across modalities. A key innovation is the Discrete Native Any-resolution Visual Transformer (dNaViT), which performs tokenization and de-tokenization at arbitrary resolutions, transforming continuous visual signals into hierarchical discrete tokens. Building on this foundation, we develop LongCat-Next, a native multimodal model that processes text, vision, and audio under a single autoregressive objective with minimal modality-specific design. As an industrial-strength foundation model, it excels at seeing, painting, and talking within a single framework, achieving strong performance across a wide range of multimodal benchmarks. In particular, LongCat-Next addresses the long-standing performance ceiling of discrete vision modeling on understanding tasks and provides a unified approach to effectively reconcile the conflict between understanding and generation. As an attempt toward native multimodality, we open-source the LongCat-Next and its tokenizers, hoping to foster further research and development in the community. GitHub: https://github.com/meituan-longcat/LongCat-Next

Amir H. Ashouri, Shayan Shirahmad Gale Bagi, Kavin Satheeskumar et al. · utoronto

The phase ordering problem has been a long-standing challenge since the late 1970s, yet it remains an open problem due to having a vast optimization space and an unbounded nature, making it an open-ended problem without a finite solution, one can limit the scope by reducing the number and the length of optimizations. Traditionally, such locally optimized decisions are made by hand-coded algorithms tuned for a small number of benchmarks, often requiring significant effort to be retuned when the benchmark suite changes. In the past 20 years, Machine Learning has been employed to construct performance models to improve the selection and ordering of compiler optimizations, however, the approaches are not baked into the compiler seamlessly and never materialized to be leveraged at a fine-grained scope of code segments. This paper presents Protean Compiler: An agile framework to enable LLVM with built-in phase-ordering capabilities at a fine-grained scope. The framework also comprises a complete library of more than 140 handcrafted static feature collection methods at varying scopes, and the experimental results showcase speedup gains of up to 4.1% on average and up to 15.7% on select Cbench applications wrt LLVM's O3 by just incurring a few extra seconds of build time on Cbench. Additionally, Protean compiler allows for an easy integration with third-party ML frameworks and other Large Language Models, and this two-step optimization shows a gain of 10.1% and 8.5% speedup wrt O3 on Cbench's Susan and Jpeg applications. Protean compiler is seamlessly integrated into LLVM and can be used as a new, enhanced, full-fledged compiler. We plan to release the project to the open-source community in the near future.

Ziwen Wang, Yancheng Yuan, Jiaming Ma et al.

In this paper, we propose a randomly projected convex clustering model for clustering a collection of $n$ high dimensional data points in $\mathbb{R}^d$ with $K$ hidden clusters. Compared to the convex clustering model for clustering original data with dimension $d$, we prove that, under some mild conditions, the perfect recovery of the cluster membership assignments of the convex clustering model, if exists, can be preserved by the randomly projected convex clustering model with embedding dimension $m = O(ε^{-2}\log(n))$, where $0 < ε< 1$ is some given parameter. We further prove that the embedding dimension can be improved to be $O(ε^{-2}\log(K))$, which is independent of the number of data points. Extensive numerical experiment results will be presented in this paper to demonstrate the robustness and superior performance of the randomly projected convex clustering model. The numerical results presented in this paper also demonstrate that the randomly projected convex clustering model can outperform the randomly projected K-means model in practice.

Yingjie Zhou, Xilei Zhu, Siyu Ren et al.

Speech-driven Talking Human (TH) generation, commonly known as "Talker," currently faces limitations in multi-subject driving capabilities. Extending this paradigm to "Multi-Talker," capable of animating multiple subjects simultaneously, introduces richer interactivity and stronger immersion in audiovisual communication. However, current Multi-Talkers still exhibit noticeable quality degradation caused by technical limitations, resulting in suboptimal user experiences. To address this challenge, we construct THQA-MT, the first large-scale Multi-Talker-generated Talking Human Quality Assessment dataset, consisting of 5,492 Multi-Talker-generated THs (MTHs) from 15 representative Multi-Talkers using 400 real portraits collected online. Through subjective experiments, we analyze perceptual discrepancies among different Multi-Talkers and identify 12 common types of distortion. Furthermore, we introduce EvalTalker, a novel TH quality assessment framework. This framework possesses the ability to perceive global quality, human characteristics, and identity consistency, while integrating Qwen-Sync to perceive multimodal synchrony. Experimental results demonstrate that EvalTalker achieves superior correlation with subjective scores, providing a robust foundation for future research on high-quality Multi-Talker generation and evaluation.

Siqi Li, Xin Li, Kunyu Yu et al.

Clinical and biomedical research in low-resource settings often faces significant challenges due to the need for high-quality data with sufficient sample sizes to construct effective models. These constraints hinder robust model training and prompt researchers to seek methods for leveraging existing knowledge from related studies to support new research efforts. Transfer learning (TL), a machine learning technique, emerges as a powerful solution by utilizing knowledge from pre-trained models to enhance the performance of new models, offering promise across various healthcare domains. Despite its conceptual origins in the 1990s, the application of TL in medical research has remained limited, especially beyond image analysis. In our review of TL applications in structured clinical and biomedical data, we screened 3,515 papers, with 55 meeting the inclusion criteria. Among these, only 2% (one out of 55) utilized external studies, and 7% (four out of 55) addressed scenarios involving multi-site collaborations with privacy constraints. To achieve actionable TL with structured medical data while addressing regional disparities, inequality, and privacy constraints in healthcare research, we advocate for the careful identification of appropriate source data and models, the selection of suitable TL frameworks, and the validation of TL models with proper baselines.

Xiao Wang, Shiao Wang, Yuhe Ding et al.

In the post-deep learning era, the Transformer architecture has demonstrated its powerful performance across pre-trained big models and various downstream tasks. However, the enormous computational demands of this architecture have deterred many researchers. To further reduce the complexity of attention models, numerous efforts have been made to design more efficient methods. Among them, the State Space Model (SSM), as a possible replacement for the self-attention based Transformer model, has drawn more and more attention in recent years. In this paper, we give the first comprehensive review of these works and also provide experimental comparisons and analysis to better demonstrate the features and advantages of SSM. Specifically, we first give a detailed description of principles to help the readers quickly capture the key ideas of SSM. After that, we dive into the reviews of existing SSMs and their various applications, including natural language processing, computer vision, graph, multi-modal and multi-media, point cloud/event stream, time series data, and other domains. In addition, we give statistical comparisons and analysis of these models and hope it helps the readers to understand the effectiveness of different structures on various tasks. Then, we propose possible research points in this direction to better promote the development of the theoretical model and application of SSM. More related works will be continuously updated on the following GitHub: https://github.com/Event-AHU/Mamba_State_Space_Model_Paper_List.

Xiao Wang, Ziwen Wang, Weizhe Kong et al.

Vehicle Re-identification (Re-ID) aims to retrieve the most similar image to a given query from images captured by non-overlapping cameras. Extending vehicle Re-ID from image-only queries to text-based queries enables retrieval in real-world scenarios where only a witness description of the target vehicle is available. In this paper, we propose PFCVR, a Part-level Fine-grained Cross-modal Vehicle Retrieval model for text-to-image vehicle re-identification. PFCVR constructs locally paired images and texts at the part level and introduces learnable part-query tokens that aggregate both part-specific and full-sentence context before aligning with visual part features. On top of this explicit local alignment, a bi-directional mask recovery module lets each modality reconstruct its masked content under the guidance of the other, implicitly bridging local correspondences into global feature alignment. Furthermore, we construct a new large-scale dataset called T2I-VeRW, which contains 14,668 images covering 1,796 vehicle identities with fine-grained part-level annotations. Experimental results on the T2I-VeRI dataset show that PFCVR achieves 29.2\% Rank-1 accuracy, improving over the best competing method by +3.7\% percentage points. On the newly proposed T2I-VeRW benchmark, PFCVR achieves 55.2\% Rank-1 accuracy, outperforming a comprehensive set of recent state-of-the-art methods. Source code will be released on https://github.com/Event-AHU/Neuromorphic_ReID

Meituan LongCat Team, Bairui Wang, Bayan et al.

We introduce LongCat-Flash-Omni, a state-of-the-art open-source omni-modal model with 560 billion parameters, excelling at real-time audio-visual interaction. By adopting a curriculum-inspired progressive training strategy that transitions from simpler to increasingly complex modality sequence modeling tasks, LongCat-Flash-Omni attains comprehensive multimodal capabilities while maintaining strong unimodal capability. Building upon LongCat-Flash, which adopts a high-performance Shortcut-connected Mixture-of-Experts (MoE) architecture with zero-computation experts, LongCat-Flash-Omni integrates efficient multimodal perception and speech reconstruction modules. Despite its immense size of 560B parameters (with 27B activated), LongCat-Flash-Omni achieves low-latency real-time audio-visual interaction. For training infrastructure, we developed a modality-decoupled parallelism scheme specifically designed to manage the data and model heterogeneity inherent in large-scale multimodal training. This innovative approach demonstrates exceptional efficiency by sustaining over 90% of the throughput achieved by text-only training. Extensive evaluations show that LongCat-Flash-Omni achieves state-of-the-art performance on omni-modal benchmarks among open-source models. Furthermore, it delivers highly competitive results across a wide range of modality-specific tasks, including text, image, and video understanding, as well as audio understanding and generation. We provide a comprehensive overview of the model architecture design, training procedures, and data strategies, and open-source the model to foster future research and development in the community.

Ziwen Wang, Stavros Birmpilis, George Labahn et al.

We describe a C implementation of the Las Vegas algorithm of Birmpilis, Labahn and Storjohann from 2020 for computing the Smith normal form of a nonsingular integer matrix. The algorithm computes a Smith massager for the input matrix using $O(n^ω\, \B(\log n + \log \|A\|)\, (\log n)^2)$ bit operations, which is softly equivalent to the cost of multiplying two matrices of the same dimension and entry size. We describe the key implementation techniques that bridge the gap between the theoretical algorithm and practical performance, including BLAS-accelerated modular arithmetic via the Residue Number System and an adaptive batching scheme that collapses the theoretical $O(\log n)$ iterations to $O(1)$ in practice. Experiments on matrices of dimension up to $n = 10007$ show that the implementation's running time scales proportionally to that of a single BLAS matrix multiplication, with both exhibiting the same effective growth rate on a log-log plot.

Junlin Liu, Shengnan An, Shuang Zhou et al.

Contemporary large language models (LLMs) have demonstrated remarkable reasoning capabilities, particularly in specialized domains like mathematics and physics. However, their ability to generalize these reasoning skills to more general and broader contexts--often termed general reasoning--remains under-explored. Unlike domain-specific reasoning, general reasoning relies less on expert knowledge but still presents formidable reasoning challenges, such as complex constraints, nested logical branches, and semantic interference. To address this gap, we introduce General365, a benchmark specifically designed to assess general reasoning in LLMs. By restricting background knowledge to a K-12 level, General365 explicitly decouples reasoning from specialized expertise. The benchmark comprises 365 seed problems and 1,095 variant problems across eight categories, ensuring both high difficulty and diversity. Evaluations across 26 leading LLMs reveal that even the top-performing model achieves only 62.8% accuracy, in stark contrast to the near-perfect performances of LLMs in math and physics benchmarks. These results suggest that the reasoning abilities of current LLMs are heavily domain-dependent, leaving significant room for improvement in broader applications. We envision General365 as a catalyst for advancing LLM reasoning beyond domain-specific tasks toward robust, general-purpose real-world scenarios. Code, Dataset, and Leaderboard: https://general365.github.io

Xiaoshan Yu, Shangshang Yang, Ziwen Wang et al.

In recent years, trustworthiness has garnered increasing attention and exploration in the field of intelligent education, due to the inherent sensitivity of educational scenarios, such as involving minors and vulnerable groups, highly personalized learning data, and high-stakes educational outcomes. However, existing research either focuses on task-specific trustworthy methods without a holistic view of trustworthy intelligent education, or provides survey-level discussions that remain high-level and fragmented, lacking a clear and systematic categorization. To address these limitations, in this paper, we present a systematic and structured review of trustworthy intelligent education. Specifically, We first organize intelligent education into five representative task categories: learner ability assessment, learning resource recommendation, learning analytics, educational content understanding, and instructional assistance. Building on this task landscape, we review existing studies from five trustworthiness perspectives, including safety and privacy, robustness, fairness, explainability, and sustainability, and summarize and categorize the research methodologies and solution strategies therein. Finally, we summarize key challenges and discuss future research directions. This survey aims to provide a coherent reference framework and facilitate a clearer understanding of trustworthiness in intelligent education.

Ziwen Wang, Shangshang Yang, Xiaoshan Yu et al.

With the advancement of network technologies, intelligent tutoring systems (ITS) have emerged to deliver increasingly precise and tailored personalized learning services. Cognitive diagnosis (CD) has emerged as a core research task in ITS, aiming to infer learners' mastery of specific knowledge concepts by modeling the mapping between learning behavior data and knowledge states. However, existing research prioritizes model performance enhancement while neglecting the pervasive noise contamination in observed response data, significantly hindering practical deployment. Furthermore, current cognitive diagnosis models (CDMs) rely heavily on researchers' domain expertise for structural design, which fails to exhaustively explore architectural possibilities, thus leaving model architectures' full potential untapped. To address this issue, we propose OSCD, an evolutionary multi-objective One-Shot neural architecture search method for Cognitive Diagnosis, designed to efficiently and robustly improve the model's capability in assessing learner proficiency. Specifically, OSCD operates through two distinct stages: training and searching. During the training stage, we construct a search space encompassing diverse architectural combinations and train a weight-sharing supernet represented via the complete binary tree topology, enabling comprehensive exploration of potential architectures beyond manual design priors. In the searching stage, we formulate the optimal architecture search under heterogeneous noise scenarios as a multi-objective optimization problem (MOP), and develop an optimization framework integrating a Pareto-optimal solution search strategy with cross-scenario performance evaluation for resolution. Extensive experiments on real-world educational datasets validate the effectiveness and robustness of the optimal architectures discovered by our OSCD model for CD tasks.

Ziwen Wang, Siqi Li, Marcus Eng Hock Ong et al.

Privacy-preserving model co-training in medical research is often hindered by server-dependent architectures incompatible with protected hospital data systems and by the predominant focus on relative effect measures (hazard ratios) which lack clinical interpretability for absolute survival risk assessment. We propose FedRD, a communication-efficient framework for federated risk difference estimation in distributed survival data. Unlike typical federated learning frameworks (e.g., FedAvg) that require persistent server connections and extensive iterative communication, FedRD is server-independent with minimal communication: one round of summary statistics exchange for the stratified model and three rounds for the unstratified model. Crucially, FedRD provides valid confidence intervals and hypothesis testing--capabilities absent in FedAvg-based frameworks. We provide theoretical guarantees by establishing the asymptotic properties of FedRD and prove that FedRD (unstratified) is asymptotically equivalent to pooled individual-level analysis. Simulation studies and real-world clinical applications across different countries demonstrate that FedRD outperforms local and federated baselines in both estimation accuracy and prediction performance, providing an architecturally feasible solution for absolute risk assessment in privacy-restricted, multi-site clinical studies.

Cheng Wang, Chen Xiong, Ziwen Wang et al.

Safety validation of autonomous driving systems requires high-risk scenario coverage, clear collision semantics, executable trajectories, and attributable multi-vehicle interactions. Existing safety-critical scenario generation methods often rely on low-level trajectory perturbations, collision-proxy optimization, or single-adversary search, which may produce adversarial samples with ambiguous collision causes or uncontrolled multi-vehicle collisions. This paper proposes KG-ASG, a collision-knowledge-guided closed-loop adversarial scenario generation framework with primary-support attribution. KG-ASG constructs a structured collision knowledge base and trains a lightweight Collision Expert to infer the target collision mode, the unique primary adversary, support vehicles, and their interaction roles. Guided by this semantic prior, multi-vehicle adversarial generation is formulated as a primary-support process, where the primary adversary induces the main conflict and support vehicles shape the surrounding risk structure without becoming additional colliders. Rule, physical, interaction-safety, and single-collider constraints are imposed as hard gates to filter non-executable samples. To handle reactive ego behaviors, planner-controller feedback is further used for failure diagnosis, candidate re-ranking, and terminal refinement. Experiments on WOMD scenarios reconstructed in MetaDrive show that KG-ASG achieves strong adversarial effectiveness while improving Valid Primary Attack, reducing multi-collision, and obtaining closed-loop recovery gains under IDM, Cruise, and Expert controllers. These results demonstrate that collision-knowledge guidance and primary-support single-collider reasoning improve adversarial effectiveness, interpretability, and executability for autonomous driving safety validation.

Meituan LongCat Team, Bayan, Bei Li et al.

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

Shangshang Yang, Mingyang Chen, Ziwen Wang et al.

Existing graph learning-based cognitive diagnosis (CD) methods have made relatively good results, but their student, exercise, and concept representations are learned and exchanged in an implicit unified graph, which makes the interaction-agnostic exercise and concept representations be learned poorly, failing to provide high robustness against noise in students' interactions. Besides, lower-order exercise latent representations obtained in shallow layers are not well explored when learning the student representation. To tackle the issues, this paper suggests a meta multigraph-assisted disentangled graph learning framework for CD (DisenGCD), which learns three types of representations on three disentangled graphs: student-exercise-concept interaction, exercise-concept relation, and concept dependency graphs, respectively. Specifically, the latter two graphs are first disentangled from the interaction graph. Then, the student representation is learned from the interaction graph by a devised meta multigraph learning module; multiple learnable propagation paths in this module enable current student latent representation to access lower-order exercise latent representations, which can lead to more effective nad robust student representations learned; the exercise and concept representations are learned on the relation and dependency graphs by graph attention modules. Finally, a novel diagnostic function is devised to handle three disentangled representations for prediction. Experiments show better performance and robustness of DisenGCD than state-of-the-art CD methods and demonstrate the effectiveness of the disentangled learning framework and meta multigraph module. The source code is available at \textcolor{red}{\url{https://github.com/BIMK/Intelligent-Education/tree/main/DisenGCD}}.

Shengnan An, Xunliang Cai, Xuezhi Cao et al.

We present AMO-Bench, an Advanced Mathematical reasoning benchmark with Olympiad level or even higher difficulty, comprising 50 human-crafted problems. Existing benchmarks have widely leveraged high school math competitions for evaluating mathematical reasoning capabilities of large language models (LLMs). However, many existing math competitions are becoming less effective for assessing top-tier LLMs due to performance saturation (e.g., AIME24/25). To address this, AMO-Bench introduces more rigorous challenges by ensuring all 50 problems are (1) cross-validated by experts to meet at least the International Mathematical Olympiad (IMO) difficulty standards, and (2) entirely original problems to prevent potential performance leakages from data memorization. Moreover, each problem in AMO-Bench requires only a final answer rather than a proof, enabling automatic and robust grading for evaluation. Experimental results across 26 LLMs on AMO-Bench show that even the best-performing model achieves only 52.4% accuracy on AMO-Bench, with most LLMs scoring below 40%. Beyond these poor performances, our further analysis reveals a promising scaling trend with increasing test-time compute on AMO-Bench. These results highlight the significant room for improving the mathematical reasoning in current LLMs. We release AMO-Bench to facilitate further research into advancing the reasoning abilities of language models. https://amo-bench.github.io/

Chen Xiong, Ziwen Wang, Deqi Wang et al.

Autonomous driving testing increasingly relies on mining safety critical scenarios from large scale naturalistic driving data, yet existing screening pipelines still depend on manual risk annotation and expensive frame by frame risk evaluation, resulting in low efficiency and weakly grounded risk quantification. To address this issue, we propose a driver risk fusion based hazardous scenario screening method for autonomous driving. During training, the method combines an improved Driver Risk Field with a dynamic cost model to generate high quality risk supervision signals, while during inference it directly predicts scenario level risk scores through fast forward passes, avoiding per frame risk computation and enabling efficient large scale ranking and retrieval. The improved Driver Risk Field introduces a new risk height function and a speed adaptive look ahead mechanism, and the dynamic cost model integrates kinetic energy, oriented bounding box constraints, and Gaussian kernel diffusion smoothing for more accurate interaction modeling. We further design a risk trajectory cross attention decoder to jointly decode risk and trajectories. Experiments on the INTERACTION and FLUID datasets show that the proposed method produces smoother and more discriminative risk estimates. On FLUID, it achieves an AUC of 0.792 and an AP of 0.825, outperforming PODAR by 9.1 percent and 5.1 percent, respectively, demonstrating its effectiveness for scalable risk labeling and hazardous scenario screening.

Junli Liang, Pengfei Zhou, Wangqiu Zhou et al.

Traditional Retrieval-Augmented Generation (RAG) effectively supports single-hop question answering with large language models but faces significant limitations in multi-hop question answering tasks, which require combining evidence from multiple documents. Existing chunk-based retrieval often provides irrelevant and logically incoherent context, leading to incomplete evidence chains and incorrect reasoning during answer generation. To address these challenges, we propose SentGraph, a sentence-level graph-based RAG framework that explicitly models fine-grained logical relationships between sentences for multi-hop question answering. Specifically, we construct a hierarchical sentence graph offline by first adapting Rhetorical Structure Theory to distinguish nucleus and satellite sentences, and then organizing them into topic-level subgraphs with cross-document entity bridges. During online retrieval, SentGraph performs graph-guided evidence selection and path expansion to retrieve fine-grained sentence-level evidence. Extensive experiments on four multi-hop question answering benchmarks demonstrate the effectiveness of SentGraph, validating the importance of explicitly modeling sentence-level logical dependencies for multi-hop reasoning.

Meituan LongCat Team, Anchun Gui, Bei Li et al.

We present LongCat-Flash-Thinking, an efficient 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model. Its advanced capabilities are cultivated through a meticulously crafted training process, beginning with long Chain-of-Thought (CoT) data cold-start and culminating in large-scale Reinforcement Learning (RL). We first employ a well-designed cold-start training strategy, which significantly enhances the reasoning potential and equips the model with specialized skills in both formal and agentic reasoning. Then, a core innovation is our domain-parallel training scheme, which decouples optimization across distinct domains (e.g., STEM, Code, Agentic) and subsequently fuses the resulting expert models into a single, nearly Pareto-optimal model. This entire process is powered by our Dynamic ORchestration for Asynchronous rollout (DORA) system, a large-scale RL framework that delivers a greater than threefold training speedup over synchronous methods on tens of thousands of accelerators. As a result, LongCat-Flash-Thinking achieves state-of-the-art performance among open-source models on a suite of complex reasoning tasks. The model exhibits exceptional efficiency in agentic reasoning, reducing average token consumption by 64.5% (from 19, 653 to 6, 965) on AIME-25, without degrading task accuracy. We release LongCat-Flash-Thinking to promote further advances in reasoning systems and agentic AI research.

Xiaolin Liu, Tianyi Zhou, Hongbo Kang et al.

Crowd evacuation simulation is critical for enhancing public safety, and demanded for realistic virtual environments. Current mainstream evacuation models overlook the complex human behaviors that occur during evacuation, such as pedestrian collisions, interpersonal interactions, and variations in behavior influenced by terrain types or individual body shapes. This results in the failure to accurately simulate the escape of people in the real world. In this paper, aligned with the sensory-decision-motor (SDM) flow of the human brain, we propose a real-time 3D crowd evacuation simulation framework that integrates a 3D-adaptive SFM (Social Force Model) Decision Mechanism and a Personalized Gait Control Motor. This framework allows multiple agents to move in parallel and is suitable for various scenarios, with dynamic crowd awareness. Additionally, we introduce Part-level Force Visualization to assist in evacuation analysis. Experimental results demonstrate that our framework supports dynamic trajectory planning and personalized behavior for each agent throughout the evacuation process, and is compatible with uneven terrain. Visually, our method generates evacuation results that are more realistic and plausible, providing enhanced insights for crowd simulation. The code is available at http://cic.tju.edu.cn/faculty/likun/projects/RESCUE.

Ziwen Wang

Single-cell RNA sequencing (scRNA-seq) has revolutionized our understanding of cellular processes by enabling gene expression analysis at the individual cell level. Clustering allows for the identification of cell types and the further discovery of intrinsic patterns in single-cell data. However, the high dimensionality and sparsity of scRNA-seq data continue to challenge existing clustering models. In this paper, we introduce JojoSCL, a novel self-supervised contrastive learning framework for scRNA-seq clustering. By incorporating a shrinkage estimator based on hierarchical Bayesian estimation, which adjusts gene expression estimates towards more reliable cluster centroids to reduce intra-cluster dispersion, and optimized using Stein's Unbiased Risk Estimate (SURE), JojoSCL refines both instance-level and cluster-level contrastive learning. Experiments on ten scRNA-seq datasets substantiate that JojoSCL consistently outperforms prevalent clustering methods, with further validation of its practicality through robustness analysis and ablation studies. JojoSCL's code is available at: https://github.com/ziwenwang28/JojoSCL.

Xiao Wang, Ziwen Wang, Wentao Wu et al.

With the rapid advancement of autonomous driving, vehicle perception, particularly detection and segmentation, has placed increasingly higher demands on algorithmic performance. Pre-trained large segmentation models, especially Segment Anything Model (SAM), have sparked significant interest and inspired new research directions in artificial intelligence. However, SAM cannot be directly applied to the fine-grained task of vehicle part segmentation, as its text-prompted segmentation functionality is not publicly accessible, and the mask regions generated by its default mode lack semantic labels, limiting its utility in structured, category-specific segmentation tasks. To address these limitations, we propose SAV, a novel framework comprising three core components: a SAM-based encoder-decoder, a vehicle part knowledge graph, and a context sample retrieval encoding module. The knowledge graph explicitly models the spatial and geometric relationships among vehicle parts through a structured ontology, effectively encoding prior structural knowledge. Meanwhile, the context retrieval module enhances segmentation by identifying and leveraging visually similar vehicle instances from training data, providing rich contextual priors for improved generalization. Furthermore, we introduce a new large-scale benchmark dataset for vehicle part segmentation, named VehicleSeg10K, which contains 11,665 high-quality pixel-level annotations across diverse scenes and viewpoints. We conduct comprehensive experiments on this dataset and two other datasets, benchmarking multiple representative baselines to establish a solid foundation for future research and comparison. % Both the dataset and source code of this paper will be released upon acceptance. Both the dataset and source code of this paper will be released on https://github.com/Event-AHU/SAV

Zhulin Jiang, Zetao Li, Cheng Wang et al.

Autonomous driving systems require comprehensive evaluation in safety-critical scenarios to ensure safety and robustness. However, such scenarios are rare and difficult to collect from real-world driving data, necessitating simulation-based synthesis. Yet, existing methods often exhibit limitations in both controllability and realism. From a capability perspective, LLMs excel at controllable generation guided by natural language instructions, while diffusion models are better suited for producing trajectories consistent with realistic driving distributions. Leveraging their complementary strengths, we propose AnchorDrive, a two-stage safety-critical scenario generation framework. In the first stage, we deploy an LLM as a driver agent within a closed-loop simulation, which reasons and iteratively outputs control commands under natural language constraints; a plan assessor reviews these commands and provides corrective feedback, enabling semantically controllable scenario generation. In the second stage, the LLM extracts key anchor points from the first-stage trajectories as guidance objectives, which jointly with other guidance terms steer the diffusion model to regenerate complete trajectories with improved realism while preserving user-specified intent. Experiments on the highD dataset demonstrate that AnchorDrive achieves superior overall performance in criticality, realism, and controllability, validating its effectiveness for generating controllable and realistic safety-critical scenarios.

Ziwen Wang, Jin Wee Lee, Tanujit Chakraborty et al.

Survival analysis is essential for studying time-to-event outcomes and providing a dynamic understanding of the probability of an event occurring over time. Various survival analysis techniques, from traditional statistical models to state-of-the-art machine learning algorithms, support healthcare intervention and policy decisions. However, there remains ongoing discussion about their comparative performance. We conducted a comparative study of several survival analysis methods, including Cox proportional hazards (CoxPH), stepwise CoxPH, elastic net penalized Cox model, Random Survival Forests (RSF), Gradient Boosting machine (GBM) learning, AutoScore-Survival, DeepSurv, time-dependent Cox model based on neural network (CoxTime), and DeepHit survival neural network. We applied the concordance index (C-index) for model goodness-of-fit, and integral Brier scores (IBS) for calibration, and considered the model interpretability. As a case study, we performed a retrospective analysis of patients admitted through the emergency department of a tertiary hospital from 2017 to 2019, predicting 90-day all-cause mortality based on patient demographics, clinicopathological features, and historical data. The results of the C-index indicate that deep learning achieved comparable performance, with DeepSurv producing the best discrimination (DeepSurv: 0.893; CoxTime: 0.892; DeepHit: 0.891). The calibration of DeepSurv (IBS: 0.041) performed the best, followed by RSF (IBS: 0.042) and GBM (IBS: 0.0421), all using the full variables. Moreover, AutoScore-Survival, using a minimal variable subset, is easy to interpret, and can achieve good discrimination and calibration (C-index: 0.867; IBS: 0.044). While all models were satisfactory, DeepSurv exhibited the best discrimination and calibration. In addition, AutoScore-Survival offers a more parsimonious model and excellent interpretability.

Ziwen Wang, Jiajun Fan, Ruihan Guo et al.

Protein generative models have shown remarkable promise in protein design but still face limitations in success rate, due to the scarcity of high-quality protein datasets for supervised pretraining. We present ProteinZero, a novel framework that enables scalable, automated, and continuous self-improvement of the inverse folding model through online reinforcement learning. To achieve computationally tractable online feedback, we introduce efficient proxy reward models based on ESM-fold and a novel rapid ddG predictor that significantly accelerates evaluation speed. ProteinZero employs a general RL framework balancing multi-reward maximization, KL-divergence from a reference model, and a novel protein-embedding level diversity regularization that prevents mode collapse while promoting higher sequence diversity. Through extensive experiments, we demonstrate that ProteinZero substantially outperforms existing methods across every key metric in protein design, achieving significant improvements in structural accuracy, designability, thermodynamic stability, and sequence diversity. Most impressively, ProteinZero reduces design failure rates by approximately 36% - 48% compared to widely-used methods like ProteinMPNN, ESM-IF and InstructPLM, consistently achieving success rates exceeding 90% across diverse and complex protein folds. Notably, the entire RL run on CATH-4.3 can be done with a single 8 X GPU node in under 3 days, including reward computation. Our work establishes a new paradigm for protein design where models evolve continuously from their own generated outputs, opening new possibilities for exploring the vast protein design space.

Zhaolu Kang, Junhao Gong, Jiaxu Yan et al.

Multimodal Large Language Models (MLLMs) have demonstrated significant potential to advance a broad range of domains. However, current benchmarks for evaluating MLLMs primarily emphasize general knowledge and vertical step-by-step reasoning typical of STEM disciplines, while overlooking the distinct needs and potential of the Humanities and Social Sciences (HSS). Tasks in the HSS domain require more horizontal, interdisciplinary thinking and a deep integration of knowledge across related fields, which presents unique challenges for MLLMs, particularly in linking abstract concepts with corresponding visual representations. Addressing this gap, we present HSSBench, a dedicated benchmark designed to assess the capabilities of MLLMs on HSS tasks in multiple languages, including the six official languages of the United Nations. We also introduce a novel data generation pipeline tailored for HSS scenarios, in which multiple domain experts and automated agents collaborate to generate and iteratively refine each sample. HSSBench contains over 13,000 meticulously designed samples, covering six key categories. We benchmark more than 20 mainstream MLLMs on HSSBench and demonstrate that it poses significant challenges even for state-of-the-art models. We hope that this benchmark will inspire further research into enhancing the cross-disciplinary reasoning abilities of MLLMs, especially their capacity to internalize and connect knowledge across fields.

Ziwen Wang, Jiajun Fan, Thao Nguyen et al.

Contrastive learning has proven to be highly efficient and adaptable in shaping representation spaces across diverse modalities by pulling similar samples together and pushing dissimilar ones apart. However, two key limitations persist: (1) Without explicit regulation of the embedding distribution, semantically related instances can inadvertently be pushed apart unless complementary signals guide pair selection, and (2) excessive reliance on large in-batch negatives and tailored augmentations hinders generalization. To address these limitations, we propose Variational Supervised Contrastive Learning (VarCon), which reformulates supervised contrastive learning as variational inference over latent class variables and maximizes a posterior-weighted evidence lower bound (ELBO) that replaces exhaustive pair-wise comparisons for efficient class-aware matching and grants fine-grained control over intra-class dispersion in the embedding space. Trained exclusively on image data, our experiments on CIFAR-10, CIFAR-100, ImageNet-100, and ImageNet-1K show that VarCon (1) achieves state-of-the-art performance for contrastive learning frameworks, reaching 79.36% Top-1 accuracy on ImageNet-1K and 78.29% on CIFAR-100 with a ResNet-50 encoder while converging in just 200 epochs; (2) yields substantially clearer decision boundaries and semantic organization in the embedding space, as evidenced by KNN classification, hierarchical clustering results, and transfer-learning assessments; and (3) demonstrates superior performance in few-shot learning than supervised baseline and superior robustness across various augmentation strategies.

Chen Chen, ZeYang Hu, Fengjiao Chen et al.

Multimodal Large Languages models have been progressing from uni-modal understanding toward unifying visual, audio and language modalities, collectively termed omni models. However, the correlation between uni-modal and omni-modal remains unclear, which requires comprehensive evaluation to drive omni model's intelligence evolution. In this work, we introduce a novel, high-quality, and UNified Omni model benchmark, UNO-Bench. This benchmark is designed to effectively evaluate both UNi-modal and Omni-modal capabilities under a unified ability taxonomy, spanning 44 task types and 5 modality combinations. It includes 1250 human curated samples for omni-modal with 98% cross-modality solvability, and 2480 enhanced uni-modal samples. The human-generated dataset is well-suited to real-world scenarios, particularly within the Chinese context, whereas the automatically compressed dataset offers a 90% increase in speed and maintains 98% consistency across 18 public benchmarks. In addition to traditional multi-choice questions, we propose an innovative multi-step open-ended question format to assess complex reasoning. A general scoring model is incorporated, supporting 6 question types for automated evaluation with 95% accuracy. Experimental result shows the Compositional Law between omni-modal and uni-modal performance and the omni-modal capability manifests as a bottleneck effect on weak models, while exhibiting synergistic promotion on strong models.

Siqi Li, Yuqing Shang, Ziwen Wang et al.

Survival analysis serves as a fundamental component in numerous healthcare applications, where the determination of the time to specific events (such as the onset of a certain disease or death) for patients is crucial for clinical decision-making. Scoring systems are widely used for swift and efficient risk prediction. However, existing methods for constructing survival scores presume that data originates from a single source, posing privacy challenges in collaborations with multiple data owners. We propose a novel framework for building federated scoring systems for multi-site survival outcomes, ensuring both privacy and communication efficiency. We applied our approach to sites with heterogeneous survival data originating from emergency departments in Singapore and the United States. Additionally, we independently developed local scores at each site. In testing datasets from each participant site, our proposed federated scoring system consistently outperformed all local models, evidenced by higher integrated area under the receiver operating characteristic curve (iAUC) values, with a maximum improvement of 11.6%. Additionally, the federated score's time-dependent AUC(t) values showed advantages over local scores, exhibiting narrower confidence intervals (CIs) across most time points. The model developed through our proposed method exhibits effective performance on each local site, signifying noteworthy implications for healthcare research. Sites participating in our proposed federated scoring model training gained benefits by acquiring survival models with enhanced prediction accuracy and efficiency. This study demonstrates the effectiveness of our privacy-preserving federated survival score generation framework and its applicability to real-world heterogeneous survival data.

Jiahong Chen, Jinghao Wang, Zi Wang et al.

6D pose estimation of textureless objects is valuable for industrial robotic applications, yet remains challenging due to the frequent loss of depth information. Current multi-view methods either rely on depth data or insufficiently exploit multi-view geometric cues, limiting their performance. In this paper, we propose DKPMV, a pipeline that achieves dense keypoint-level fusion using only multi-view RGB images as input. We design a three-stage progressive pose optimization strategy that leverages dense multi-view keypoint geometry information. To enable effective dense keypoint fusion, we enhance the keypoint network with attentional aggregation and symmetry-aware training, improving prediction accuracy and resolving ambiguities on symmetric objects. Extensive experiments on the ROBI dataset demonstrate that DKPMV outperforms state-of-the-art multi-view RGB approaches and even surpasses the RGB-D methods in the majority of cases. The code will be available soon.

Shangshang Yang, Jialin Han, Xiaoshan Yu et al.

Online intelligent education platforms have generated a vast amount of distributed student learning data. This influx of data presents opportunities for cognitive diagnosis (CD) to assess students' mastery of knowledge concepts while also raising significant data privacy and security challenges. To cope with this issue, federated learning (FL) becomes a promising solution by jointly training models across multiple local clients without sharing their original data. However, the data quality problem, caused by the ability differences and educational context differences between different groups/schools of students, further poses a challenge to the fairness of models. To address this challenge, this paper proposes a fairness-aware federated cognitive diagnosis framework (FedCD) to jointly train CD models built upon a novel parameter decoupling-based personalization strategy, preserving privacy of data and achieving precise and fair diagnosis of students on each client. As an FL paradigm, FedCD trains a local CD model for the students in each client based on its local student learning data, and each client uploads its partial model parameters to the central server for parameter aggregation according to the devised innovative personalization strategy. The main idea of this strategy is to decouple model parameters into two parts: the first is used as locally personalized parameters, containing diagnostic function-related model parameters, to diagnose each client's students fairly; the second is the globally shared parameters across clients and the server, containing exercise embedding parameters, which are updated via fairness-aware aggregation, to alleviate inter-school unfairness. Experiments on three real-world datasets demonstrate the effectiveness of the proposed FedCD framework and the personalization strategy compared to five FL approaches under three CD models.

Amir H. Ashouri, Muhammad Asif Manzoor, Duc Minh Vu et al. · utoronto

The key to performance optimization of a program is to decide correctly when a certain transformation should be applied by a compiler. This is an ideal opportunity to apply machine-learning models to speed up the tuning process; while this realization has been around since the late 90s, only recent advancements in ML enabled a practical application of ML to compilers as an end-to-end framework. This paper presents ACPO: An AI-Enabled Compiler Framework, a novel framework that provides LLVM with simple and comprehensive tools to benefit from employing ML models for different optimization passes. We first showcase the high-level view, class hierarchy, and functionalities of ACPO and subsequently, demonstrate \taco{a couple of use cases of ACPO by ML-enabling the Loop Unroll and Function Inlining passes used in LLVM's O3. and finally, describe how ACPO can be leveraged to optimize other passes. Experimental results reveal that the ACPO model for Loop Unroll can gain on average 4%, 3%, 5.4%, and 0.2% compared to LLVM's vanilla O3 optimization when deployed on Polybench, Coral-2, CoreMark, and Graph-500, respectively. Furthermore, by including both Function Inlining and Loop Unroll models, ACPO can provide a combined speedup of 4.5% on Polybench and 2.4% on Cbench when compared with LLVM's O3, respectively.

Henry H. Yu, Xue Feng, Hao Sun et al.

Convolutional neural networks (CNNs) have been successfully applied to medical image classification, segmentation, and related tasks. Among the many CNNs architectures, U-Net and its improved versions based are widely used and achieve state-of-the-art performance these years. These improved architectures focus on structural improvements and the size of the convolution kernel is generally fixed. In this paper, we propose a module that combines the benefits of multiple kernel sizes and we apply the proposed module to U-Net and its variants. We test our module on three segmentation benchmark datasets and experimental results show significant improvement.

Henry H. Yu, Jiang Liu, Hao Sun et al.

Image pairing is an important research task in the field of computer vision. And finding image pairs containing objects of the same category is the basis of many tasks such as tracking and person re-identification, etc., and it is also the focus of our research. Existing traditional methods and deep learning-based methods have some degree of defects in speed or accuracy. In this paper, we made improvements on the Siamese network and proposed GetNet. The proposed method GetNet combines STN and Siamese network to get the target area first and then perform subsequent processing. Experiments show that our method achieves competitive results in speed and accuracy.

Haiqian Gu, Jie Wang, Ziwen Wang et al.

With the booming development and popularity of mobile applications, different verticals accumulate abundant data of user information and social behavior, which are spontaneous, genuine and diversified. However, each platform describes user's portraits in only certain aspect, resulting in difficult combination of those internet footprints together. In our research, we proposed a modeling approach to analyze user's online behavior across different social media platforms. Structured and unstructured data of same users shared by NetEase Music and Sina Weibo have been collected for cross-platform analysis of correlations between music preference and other users' characteristics. Based on music tags of genre and mood, genre cluster of five groups and mood cluster of four groups have been formed by computing their collected song lists with K-means method. Moreover, with the help of user data of Weibo, correlations between music preference (i.e. genre, mood) and Big Five personalities (BFPs) and basic information (e.g. gender, resident region, tags) have been comprehensively studied, building up full-scale user portraits with finer grain. Our findings indicate that people's music preference could be linked with their real social activities. For instance, people living in mountainous areas generally prefer folk music, while those in urban areas like pop music more. Interestingly, dog lovers could love sad music more than cat lovers. Moreover, our proposed cross-platform modeling approach could be adapted to other verticals, providing an online automatic way for profiling users in a more precise and comprehensive way.

Ziwen Wang, Jie Wang, Haiqian Gu et al.

Automatic text generation has received much attention owing to rapid development of deep neural networks. In general, text generation systems based on statistical language model will not consider anthropomorphic characteristics, which results in machine-like generated texts. To fill the gap, we propose a conditional language generation model with Big Five Personality (BFP) feature vectors as input context, which writes human-like short texts. The short text generator consists of a layer of long short memory network (LSTM), where a BFP feature vector is concatenated as one part of input for each cell. To enable supervised training generation model, a text classification model based convolution neural network (CNN) has been used to prepare BFP-tagged Chinese micro-blog corpora. Validated by a BFP linguistic computational model, our generated Chinese short texts exhibit discriminative personality styles, which are also syntactically correct and semantically smooth with appropriate emoticons. With combination of natural language generation with psychological linguistics, our proposed BFP-dependent text generation model can be widely used for individualization in machine translation, image caption, dialogue generation and so on.