Jin Zhang

Weiwei Ding, Zixuan Li, Long Bai et al.

Knowledge Graphs (KGs) are widely used to mitigate the limitations of Large Language Models (LLMs), such as outdated knowledge and hallucinations. Existing LLM-KG integration frameworks typically rely on predefined operators to retrieve factual knowledge from KGs and inject it into prompts for answer generation. This paradigm faces two critical bottlenecks: 1) Inflexibility: The predefined operators are limited in scope and thus lack sufficient compositional expressiveness to fully capture the complex semantics required by KG questions. 2) Unscalability: Direct injection of factual knowledge into prompts limits scalability in handling large-scale factual knowledge. To address these two bottlenecks, we propose Code-on-Graph (CoG), a programmatic reasoning framework for LLM-KG integration. Specifically, given the factual knowledge retrieved at each reasoning step, CoG first identifies the corresponding KG schemas and represents these schemas as Python classes, which serve as abstract interfaces to the retrieved facts. It then generates executable code grounded in these classes, with the retrieved facts instantiated as objects of the corresponding classes during execution. This design enables flexible code-based reasoning while avoiding the direct injection of large-scale factual knowledge into prompts. Experiments on WebQSP, CWQ, and GrailQA demonstrate that CoG outperforms prior state-of-the-art models by up to 10.5%.

Kimi Team, Tongtong Bai, Yifan Bai et al.

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

Yicheng Zou, Dongsheng Zhu, Lin Zhu et al.

We introduce Intern-S1-Pro, the first one-trillion-parameter scientific multimodal foundation model. Scaling to this unprecedented size, the model delivers a comprehensive enhancement across both general and scientific domains. Beyond stronger reasoning and image-text understanding capabilities, its intelligence is augmented with advanced agent capabilities. Simultaneously, its scientific expertise has been vastly expanded to master over 100 specialized tasks across critical science fields, including chemistry, materials, life sciences, and earth sciences. Achieving this massive scale is made possible by the robust infrastructure support of XTuner and LMDeploy, which facilitates highly efficient Reinforcement Learning (RL) training at the 1-trillion parameter level while ensuring strict precision consistency between training and inference. By seamlessly integrating these advancements, Intern-S1-Pro further fortifies the fusion of general and specialized intelligence, working as a Specializable Generalist, demonstrating its position in the top tier of open-source models for general capabilities, while outperforming proprietary models in the depth of specialized scientific tasks.

Andrey Ignatov, Radu Timofte, Maurizio Denna et al.

Image super-resolution is a common task on mobile and IoT devices, where one often needs to upscale and enhance low-resolution images and video frames. While numerous solutions have been proposed for this problem in the past, they are usually not compatible with low-power mobile NPUs having many computational and memory constraints. In this Mobile AI challenge, we address this problem and propose the participants to design an efficient quantized image super-resolution solution that can demonstrate a real-time performance on mobile NPUs. The participants were provided with the DIV2K dataset and trained INT8 models to do a high-quality 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated edge NPU capable of accelerating quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 60 FPS rate when reconstructing Full HD resolution images. A detailed description of all models developed in the challenge is provided in this paper.

Eduardo Pérez-Pellitero, Sibi Catley-Chandar, Richard Shaw et al.

This paper reviews the challenge on constrained high dynamic range (HDR) imaging that was part of the New Trends in Image Restoration and Enhancement (NTIRE) workshop, held in conjunction with CVPR 2022. This manuscript focuses on the competition set-up, datasets, the proposed methods and their results. The challenge aims at estimating an HDR image from multiple respective low dynamic range (LDR) observations, which might suffer from under- or over-exposed regions and different sources of noise. The challenge is composed of two tracks with an emphasis on fidelity and complexity constraints: In Track 1, participants are asked to optimize objective fidelity scores while imposing a low-complexity constraint (i.e. solutions can not exceed a given number of operations). In Track 2, participants are asked to minimize the complexity of their solutions while imposing a constraint on fidelity scores (i.e. solutions are required to obtain a higher fidelity score than the prescribed baseline). Both tracks use the same data and metrics: Fidelity is measured by means of PSNR with respect to a ground-truth HDR image (computed both directly and with a canonical tonemapping operation), while complexity metrics include the number of Multiply-Accumulate (MAC) operations and runtime (in seconds).

Andrey Ignatov, Radu Timofte, Jin Zhang et al.

As mobile cameras with compact optics are unable to produce a strong bokeh effect, lots of interest is now devoted to deep learning-based solutions for this task. In this Mobile AI challenge, the target was to develop an efficient end-to-end AI-based bokeh effect rendering approach that can run on modern smartphone GPUs using TensorFlow Lite. The participants were provided with a large-scale EBB! bokeh dataset consisting of 5K shallow / wide depth-of-field image pairs captured using the Canon 7D DSLR camera. The runtime of the resulting models was evaluated on the Kirin 9000's Mali GPU that provides excellent acceleration results for the majority of common deep learning ops. A detailed description of all models developed in this challenge is provided in this paper.

Jin Zhang, Qiuwei Liang, Yanjiao Shi

Most existing salient object detection (SOD) models are difficult to apply due to the complex and huge model structures. Although some lightweight models are proposed, the accuracy is barely satisfactory. In this paper, we design a novel semantics-guided contextual fusion network (SCFNet) that focuses on the interactive fusion of multi-level features for accurate and efficient salient object detection. Furthermore, we apply knowledge distillation to SOD task and provide a sizeable dataset KD-SOD80K. In detail, we transfer the rich knowledge from a seasoned teacher to the untrained SCFNet through unlabeled images, enabling SCFNet to learn a strong generalization ability to detect salient objects more accurately. The knowledge distillation based SCFNet (KDSCFNet) achieves comparable accuracy to the state-of-the-art heavyweight methods with less than 1M parameters and 174 FPS real-time detection speed. Extensive experiments demonstrate the robustness and effectiveness of the proposed distillation method and SOD framework. Code and data: https://github.com/zhangjinCV/KD-SCFNet.

Risheng Liu, Yaohua Liu, Wei Yao et al.

Gradient methods have become mainstream techniques for Bi-Level Optimization (BLO) in learning fields. The validity of existing works heavily rely on either a restrictive Lower-Level Strong Convexity (LLSC) condition or on solving a series of approximation subproblems with high accuracy or both. In this work, by averaging the upper and lower level objectives, we propose a single loop Bi-level Averaged Method of Multipliers (sl-BAMM) for BLO that is simple yet efficient for large-scale BLO and gets rid of the limited LLSC restriction. We further provide non-asymptotic convergence analysis of sl-BAMM towards KKT stationary points, and the comparative advantage of our analysis lies in the absence of strong gradient boundedness assumption, which is always required by others. Thus our theory safely captures a wider variety of applications in deep learning, especially where the upper-level objective is quadratic w.r.t. the lower-level variable. Experimental results demonstrate the superiority of our method.

Siyuan Li, Hao Li, Jin Zhang et al. · tsinghua

Humans have the ability to reuse previously learned policies to solve new tasks quickly, and reinforcement learning (RL) agents can do the same by transferring knowledge from source policies to a related target task. Transfer RL methods can reshape the policy optimization objective (optimization transfer) or influence the behavior policy (behavior transfer) using source policies. However, selecting the appropriate source policy with limited samples to guide target policy learning has been a challenge. Previous methods introduce additional components, such as hierarchical policies or estimations of source policies' value functions, which can lead to non-stationary policy optimization or heavy sampling costs, diminishing transfer effectiveness. To address this challenge, we propose a novel transfer RL method that selects the source policy without training extra components. Our method utilizes the Q function in the actor-critic framework to guide policy selection, choosing the source policy with the largest one-step improvement over the current target policy. We integrate optimization transfer and behavior transfer (IOB) by regularizing the learned policy to mimic the guidance policy and combining them as the behavior policy. This integration significantly enhances transfer effectiveness, surpasses state-of-the-art transfer RL baselines in benchmark tasks, and improves final performance and knowledge transferability in continual learning scenarios. Additionally, we show that our optimization transfer technique is guaranteed to improve target policy learning.

Risheng Liu, Xuan Liu, Shangzhi Zeng et al.

In recent years, by utilizing optimization techniques to formulate the propagation of deep model, a variety of so-called Optimization-Derived Learning (ODL) approaches have been proposed to address diverse learning and vision tasks. Although having achieved relatively satisfying practical performance, there still exist fundamental issues in existing ODL methods. In particular, current ODL methods tend to consider model construction and learning as two separate phases, and thus fail to formulate their underlying coupling and depending relationship. In this work, we first establish a new framework, named Hierarchical ODL (HODL), to simultaneously investigate the intrinsic behaviors of optimization-derived model construction and its corresponding learning process. Then we rigorously prove the joint convergence of these two sub-tasks, from the perspectives of both approximation quality and stationary analysis. To our best knowledge, this is the first theoretical guarantee for these two coupled ODL components: optimization and learning. We further demonstrate the flexibility of our framework by applying HODL to challenging learning tasks, which have not been properly addressed by existing ODL methods. Finally, we conduct extensive experiments on both synthetic data and real applications in vision and other learning tasks to verify the theoretical properties and practical performance of HODL in various application scenarios.

Risheng Liu, Xuan Liu, Shangzhi Zeng et al.

Recently, Optimization-Derived Learning (ODL) has attracted attention from learning and vision areas, which designs learning models from the perspective of optimization. However, previous ODL approaches regard the training and hyper-training procedures as two separated stages, meaning that the hyper-training variables have to be fixed during the training process, and thus it is also impossible to simultaneously obtain the convergence of training and hyper-training variables. In this work, we design a Generalized Krasnoselskii-Mann (GKM) scheme based on fixed-point iterations as our fundamental ODL module, which unifies existing ODL methods as special cases. Under the GKM scheme, a Bilevel Meta Optimization (BMO) algorithmic framework is constructed to solve the optimal training and hyper-training variables together. We rigorously prove the essential joint convergence of the fixed-point iteration for training and the process of optimizing hyper-parameters for hyper-training, both on the approximation quality, and on the stationary analysis. Experiments demonstrate the efficiency of BMO with competitive performance on sparse coding and real-world applications such as image deconvolution and rain streak removal.

Risheng Liu, Xuan Liu, Wei Yao et al.

Gradient methods have become mainstream techniques for Bi-Level Optimization (BLO) in learning and vision fields. The validity of existing works heavily relies on solving a series of approximation subproblems with extraordinarily high accuracy. Unfortunately, to achieve the approximation accuracy requires executing a large quantity of time-consuming iterations and computational burden is naturally caused. This paper is thus devoted to address this critical computational issue. In particular, we propose a single-level formulation to uniformly understand existing explicit and implicit Gradient-based BLOs (GBLOs). This together with our designed counter-example can clearly illustrate the fundamental numerical and theoretical issues of GBLOs and their naive accelerations. By introducing the dual multipliers as a new variable, we then establish Bilevel Alternating Gradient with Dual Correction (BAGDC), a general framework, which significantly accelerates different categories of existing methods by taking specific settings. A striking feature of our convergence result is that, compared to those original unaccelerated GBLO versions, the fast BAGDC admits a unified non-asymptotic convergence theory towards stationarity. A variety of numerical experiments have also been conducted to demonstrate the superiority of the proposed algorithmic framework.

Jin Zhang, Siyuan Li, Chongjie Zhang

The ability to reuse previous policies is an important aspect of human intelligence. To achieve efficient policy reuse, a Deep Reinforcement Learning (DRL) agent needs to decide when to reuse and which source policies to reuse. Previous methods solve this problem by introducing extra components to the underlying algorithm, such as hierarchical high-level policies over source policies, or estimations of source policies' value functions on the target task. However, training these components induces either optimization non-stationarity or heavy sampling cost, significantly impairing the effectiveness of transfer. To tackle this problem, we propose a novel policy reuse algorithm called Critic-gUided Policy reuse (CUP), which avoids training any extra components and efficiently reuses source policies. CUP utilizes the critic, a common component in actor-critic methods, to evaluate and choose source policies. At each state, CUP chooses the source policy that has the largest one-step improvement over the current target policy, and forms a guidance policy. The guidance policy is theoretically guaranteed to be a monotonic improvement over the current target policy. Then the target policy is regularized to imitate the guidance policy to perform efficient policy search. Empirical results demonstrate that CUP achieves efficient transfer and significantly outperforms baseline algorithms.

Lucy Gao, Jane J. Ye, Haian Yin et al.

Gradient-based optimization methods for hyperparameter tuning guarantee theoretical convergence to stationary solutions when for fixed upper-level variable values, the lower level of the bilevel program is strongly convex (LLSC) and smooth (LLS). This condition is not satisfied for bilevel programs arising from tuning hyperparameters in many machine learning algorithms. In this work, we develop a sequentially convergent Value Function based Difference-of-Convex Algorithm with inexactness (VF-iDCA). We show that this algorithm achieves stationary solutions without LLSC and LLS assumptions for bilevel programs from a broad class of hyperparameter tuning applications. Our extensive experiments confirm our theoretical findings and show that the proposed VF-iDCA yields superior performance when applied to tune hyperparameters.

Dachong Li, ZhuangZhuang Chen, Jin Zhang et al.

Vision--Language--Action (VLA) models often use intermediate representations to connect multimodal inputs with continuous control, yet spatial guidance is often injected implicitly through latent features. We propose $CorridorVLA$, which predicts sparse spatial anchors as incremental physical changes (e.g., $Δ$-positions) and uses them to impose an explicit tolerance region in the training objective for action generation. The anchors define a corridor that guides a flow-matching action head: trajectories whose implied spatial evolution falls outside it receive corrective gradients, while minor deviations from contacts and execution noise are permitted. On the more challenging LIBERO-Plus benchmark, CorridorVLA yields consistent gains across both SmolVLA and GR00T, improving success rate by $3.4\%$--$12.4\%$ over the corresponding baselines; notably, our GR00T-Corr variant reaches a success rate of $83.21\%$. These results indicate that action-aligned physical cues can provide direct and interpretable constraints for generative action policies, complementing spatial guidance encoded in visual or latent forms. Code is available at https://github.com/corridorVLA.

Jin Zhang, Xiaowei Liu

In this paper, we analyze the supercloseness property of the streamline diffusion finite element method (SDFEM) on Shishkin triangular meshes, which is different from one in the case of rectangular meshes. The analysis depends on integral inequalities for the part related to the diffusion in the bilinear form. Moreover, our result allows the construction of a simple postprocessing that yields a more accurate solution. Finally, numerical experiments support these theoretical results.

Lucy L. Gao, Jane J. Ye, Haian Yin et al.

Bilevel programming has emerged as a valuable tool for hyperparameter selection, a central concern in machine learning. In a recent study by Ye et al. (2023), a value function-based difference of convex algorithm was introduced to address bilevel programs. This approach proves particularly powerful when dealing with scenarios where the lower-level problem exhibits convexity in both the upper-level and lower-level variables. Examples of such scenarios include support vector machines and $\ell_1$ and $\ell_2$ regularized regression. In this paper, we significantly expand the range of applications, now requiring convexity only in the lower-level variables of the lower-level program. We present an innovative single-level difference of weakly convex reformulation based on the Moreau envelope of the lower-level problem. We further develop a sequentially convergent Inexact Proximal Difference of Weakly Convex Algorithm (iP-DwCA). To evaluate the effectiveness of the proposed iP-DwCA, we conduct numerical experiments focused on tuning hyperparameters for kernel support vector machines on simulated data.

Jin Zhang, Xiaowei Liu

In this paper, we analyze the streamline diffusion finite element method (SDFEM) for a model singularly perturbed convection-diffusion equation on a Shishkin triangular mesh and hybrid meshes. Supercloseness property of $u^I-u^N$ is obtained, where $u^I$ is the interpolant of the solution $u$ and $u^N$ is the SDFEM's solution. The analysis depends on novel integral inequalities for the diffusion and convection parts in the bilinear form. Furthermore, analysis on hybrid meshes shows that bilinear elements should be recommended for the exponential layer, not for the characteristic layer. Finally, numerical experiments support these theoretical results.

Linling Kuang, Jiachen Sun, Jin Zhang et al.

As one of the most promising hotspots in the 6G era, space remote sensing information networks play a key and irreplaceable role in areas such as emergency response and scientific research, and are expected to foster remote sensing data processing into the next generation of killer applications. However, due to the inability to deploy ground communication stations at scale and the limited satellite-to-ground link rate, the traditional model for transmitting space data back to ground stations faces significant challenges in terms of timeliness. To address this problem, we focus on the emerging paradigm of on-orbit space data processing, which reduces the volume of transmitted data by several orders of magnitude to enable faster task response, taking the first step toward building a space-based computing network. Specifically, we propose a hierarchical space-based computing network architecture, comprising the space-based cloud constellation system, the remote sensing constellation system, the network operation control center, the orchestration data center, and the user access portal. Each component is described in detail from a system design perspective to clarify its specific role and functionality. Next, we analyze three scientific challenges: the heterogeneous resource virtualization and state information synchronization, the matching of multi-priority tasks with multidimensional resources, and the fault detection and localization under extreme conditions. Finally, we discuss key technologies to address the aforementioned challenges and highlight promising research priorities for the future.

Yanglong Song, Zihao Yang, Shuo Meng et al.

Generative sequence models have shown strong results in recommendation. Applying them to search ranking is more challenging. Search behavior is inherently query-driven. Each query switch introduces a sharp topic shift in the user's interaction history. Existing generative methods flatten queries and items into a single token sequence. They do not distinguish query boundaries. This causes the model to mix different query intents into one prediction target, resulting in noisy supervision. We present Query-Conditioned Generative Search (QGS). QGS encodes each interaction as a (query, item) pair token. It trains with a query-conditioned next-item objective. The prediction target changes from a noisy marginal P(item_{t+1}|context_{<=t}) to a clean conditional P(item_{t+1}|context_{<=t}, query_{t+1}). This directly removes the semantic discontinuity caused by query switches. Encoding long interaction histories with standard attention has quadratic cost. This is impractical under strict online latency budgets. We introduce a Linear HSTU encoder. It replaces full attention with causal linear recurrence. Per-layer complexity drops from O(L^2) to O(L) with no loss in ranking quality. Traditional search ranking depends on hand-crafted features like text-matching scores, statistical signals, and behavioral features. We propose HFG-Attention to preserve them in the generative framework. It organizes heterogeneous features into semantic groups and fuses them through a dedicated attention block. This bridges sparse engineered signals with dense sequential representations. QGS is deployed in the ranking module of Quark Search, a major commercial search engine in China. Online A/B tests show statistically significant gains: +0.62% CTR, +0.38% Click-Search Ratio, and +3.55% PV Duration over the production deep learning baseline.

JiaXu Wan, Hong Zhang, Jin Zhang et al.

Many multi-object tracking (MOT) methods follow the framework of "tracking by detection", which associates the target objects-of-interest based on the detection results. However, due to the separate models for detection and association, the tracking results are not optimal.Moreover, the speed is limited by some cumbersome association methods to achieve high tracking performance. In this work, we propose an end-to-end MOT method, with a Gaussian filter-inspired dynamic search region refinement module to dynamically filter and refine the search region by considering both the template information from the past frames and the detection results from the current frame with little computational burden, and a lightweight attention-based tracking head to achieve the effective fine-grained instance association. Extensive experiments and ablation study on MOT17 and MOT20 datasets demonstrate that our method can achieve the state-of-the-art performance with reasonable speed.

Jin Zhang

We propose some useful estimates for the pointwise error estimates of the streamline diffusion finite element method (SDFEM) on Shishkin meshes, when SDFEM is applied for problems of characteristic layers.

Jin Zhang, Jiachen Sun, Kai Liu et al.

As industry and academia continue to advance spaceborne computing and communication capabilities, the formation of cloud-native space clusters (CNSCs) has become an increasingly evident trend. This evolution progressively exposes the resource management challenges associated with coordinating fragmented and heterogeneous onboard resources while supporting large-scale and diverse space applications. However, directly transplanting mature terrestrial cloud-native cluster operating system paradigms into space is ineffective due to the fragmentation of spaceborne computing resources and satellite mobility, which collectively impose substantial challenges on resource awareness and orchestration. This article presents YUHENG-OS, a cloud-native space cluster operating system tailored for CNSCs. YUHENG-OS provides unified abstraction, awareness, and orchestration of heterogeneous spaceborne infrastructure, enabling cluster-wide task deployment and scheduling across distributed satellites. We introduce a four-layer system architecture and three key enabling technologies: modeling of heterogeneous resource demands for space tasks, fragmented heterogeneous resource awareness under network constraints, and matching of differentiated tasks with multidimensional heterogeneous resources under temporal dependency constraints. Evaluation results show that, compared with representative terrestrial cloud-native cluster operating systems exemplified by Kubernetes, YUHENG-OS achieves a substantially higher task completion ratio, with improvements of up to 98%. This advantage is primarily attributed to its ability to reduce resource awareness delay by 71%.

Yufeng Wang, Cong Xu, Min Yang et al.

Although Physics-Informed Neural Networks (PINNs) have been successfully applied in a wide variety of science and engineering fields, they can fail to accurately predict the underlying solution in slightly challenging convection-diffusion-reaction problems. In this paper, we investigate the reason of this failure from a domain distribution perspective, and identify that learning multi-scale fields simultaneously makes the network unable to advance its training and easily get stuck in poor local minima. We show that the widespread experience of sampling more collocation points in high-loss layer regions hardly help optimize and may even worsen the results. These findings motivate the development of a novel curriculum learning method that encourages neural networks to prioritize learning on easier non-layer regions while downplaying learning on harder layer regions. The proposed method helps PINNs automatically adjust the learning emphasis and thereby facilitate the optimization procedure. Numerical results on typical benchmark equations show that the proposed curriculum learning approach mitigates the failure modes of PINNs and can produce accurate results for very sharp boundary and interior layers. Our work reveals that for equations whose solutions have large scale differences, paying less attention to high-loss regions can be an effective strategy for learning them accurately.

Kabir Swain, Sijie Han, Ayush Raina et al.

As large language models (LLMs) continue to improve in reasoning and decision-making, there is a growing need for realistic and interactive environments where their abilities can be rigorously evaluated. We present VirtualEnv, a next-generation simulation platform built on Unreal Engine 5 that enables fine-grained benchmarking of LLMs in embodied and interactive scenarios. VirtualEnv supports rich agent-environment interactions, including object manipulation, navigation, and adaptive multi-agent collaboration, as well as game-inspired mechanics like escape rooms and procedurally generated environments. We provide a user-friendly API built on top of Unreal Engine, allowing researchers to deploy and control LLM-driven agents using natural language instructions. We integrate large-scale LLMs and vision-language models (VLMs), such as GPT-based models, to generate novel environments and structured tasks from multimodal inputs. Our experiments benchmark the performance of several popular LLMs across tasks of increasing complexity, analyzing differences in adaptability, planning, and multi-agent coordination. We also describe our methodology for procedural task generation, task validation, and real-time environment control. VirtualEnv is released as an open-source platform, we aim to advance research at the intersection of AI and gaming, enable standardized evaluation of LLMs in embodied AI settings, and pave the way for future developments in immersive simulations and interactive entertainment.

Xiaofei Huang, Hongfang Gong, Jin Zhang

The Transformer has been successfully used in medical image segmentation due to its excellent long-range modeling capabilities. However, patch segmentation is necessary when building a Transformer class model. This process may disrupt the tissue structure in medical images, resulting in the loss of relevant information. In this study, we proposed a Heterogeneous Swin Transformer with Multi-Receptive Field (HST-MRF) model based on U-shaped networks for medical image segmentation. The main purpose is to solve the problem of loss of structural information caused by patch segmentation using transformer by fusing patch information under different receptive fields. The heterogeneous Swin Transformer (HST) is the core module, which achieves the interaction of multi-receptive field patch information through heterogeneous attention and passes it to the next stage for progressive learning. We also designed a two-stage fusion module, multimodal bilinear pooling (MBP), to assist HST in further fusing multi-receptive field information and combining low-level and high-level semantic information for accurate localization of lesion regions. In addition, we developed adaptive patch embedding (APE) and soft channel attention (SCA) modules to retain more valuable information when acquiring patch embedding and filtering channel features, respectively, thereby improving model segmentation quality. We evaluated HST-MRF on multiple datasets for polyp and skin lesion segmentation tasks. Experimental results show that our proposed method outperforms state-of-the-art models and can achieve superior performance. Furthermore, we verified the effectiveness of each module and the benefits of multi-receptive field segmentation in reducing the loss of structural information through ablation experiments.

Xiang Li, Shuwei Chen, Jian Dong et al.

Click-through rate (CTR) prediction is crucial in recommendation and online advertising systems. Existing methods usually model user behaviors, while ignoring the informative context which influences the user to make a click decision, e.g., click pages and pre-ranking candidates that inform inferences about user interests, leading to suboptimal performance. In this paper, we propose a Decision-Making Context Interaction Network (DCIN), which deploys a carefully designed Context Interaction Unit (CIU) to learn decision-making contexts and thus benefits CTR prediction. In addition, the relationship between different decision-making context sources is explored by the proposed Adaptive Interest Aggregation Unit (AIAU) to improve CTR prediction further. In the experiments on public and industrial datasets, DCIN significantly outperforms the state-of-the-art methods. Notably, the model has obtained the improvement of CTR+2.9%/CPM+2.1%/GMV+1.5% for online A/B testing and served the main traffic of Meituan Waimai advertising system.

Jin Zhang, Ruiheng Zhang, Yanjiao Shi et al.

Existing Camouflaged Object Detection (COD) methods rely heavily on large-scale pixel-annotated training sets, which are both time-consuming and labor-intensive. Although weakly supervised methods offer higher annotation efficiency, their performance is far behind due to the unclear visual demarcations between foreground and background in camouflaged images. In this paper, we explore the potential of using boxes as prompts in camouflaged scenes and introduce the first weakly semi-supervised COD method, aiming for budget-efficient and high-precision camouflaged object segmentation with an extremely limited number of fully labeled images. Critically, learning from such limited set inevitably generates pseudo labels with serious noisy pixels. To address this, we propose a noise correction loss that facilitates the model's learning of correct pixels in the early learning stage, and corrects the error risk gradients dominated by noisy pixels in the memorization stage, ultimately achieving accurate segmentation of camouflaged objects from noisy labels. When using only 20% of fully labeled data, our method shows superior performance over the state-of-the-art methods.

Guikang Du, Haoran Li, Xinyu Liu et al.

Cross-subject generalization in biomedical time-series refers to training on data from some subjects and testing on unseen subjects.The key challenge is to suppress subject specific variability in BTS representations.Most existing methods implicitly suppress the variability through model building or subject adversarial learning, but rarely model it explicitly.We introduce spectral drift as a new perspective to characterize subject specific variability.Specifically, BTS signals under the same label often share consistent oscillatory structure, yet exhibit subject-dependent magnitude or phase shifts in specific frequency components, which we interpret as subject-specific variability. Building on this insight, we propose BioFormer.At its core is a Frequency-Band Alignment Module(FBAM) that generates band-wise modulation factors from the spectral distribution and adaptively adjusts amplitude and phase to align spectral structure, thereby mitigating variability.We further pair FBAM with Sample Conditional Layer Normalization, which infers normalization parameters from intrinsic signal statistics rather than subject identity, stabilizing cross-subject representations.Extensive experiments on six datasets demonstrate that BioFormer outperforms 12 baselines, yielding absolute F1-score improvements of 6%.

Jin Zhang

In this technical report, we present estimations of the discrete Green's function of the streamline diffusion finite element method (SDFEM) on Shishkin triangular meshes for problems with only exponential layers.

Wenqiang Huang, Susu Fang, Xuhang Gu et al.

Exemplified by the chemical vapor deposition growth of two-dimensional dendrites, which has potential applications in catalysis and presents a parameter-intensive, data-scarce and reaction process-complex model problem, we devise a machine intelligence-empowered framework for the full chain support of material synthesis, encompassing rapid process optimization, accurate customized synthesis, and comprehensive mechanism deciphering.First, active learning is integrated into the experimental workflow, identifying an optimal recipe for the growth of highly-branched, electrocatalytically-active ReSe2 dendrites through 60 experiments (4 iterations), which account for less than 1.3% of the numerous possible parameter combinations.Then, a prediction accuracy-guided data augmentation strategy is developed combined with a tree-based machine learning (ML) algorithm, unveiling a non-linear correlation between 5 process variables and fractal dimension (DF) of ReSe2 dendrites with only 9 experiment additions, which guides the synthesis of various user-defined DF. Finally, we construct a data-knowledge dual-driven mechanism model by integration of cross-scale characterizations, interpretable ML models, and domain knowledge in thermodynamics and kinetics, unraveling synergistic contributions of multiple process parameters to the product morphology. This work demonstrates the ML potential to transform the research paradigm and is adaptable to broader material synthesis.

Lei Bai, Zhongrui Cai, Yuhang Cao et al.

In recent years, a plethora of open-source foundation models have emerged, achieving remarkable progress in some widely attended fields, with performance being quite close to that of closed-source models. However, in high-value but more challenging scientific professional fields, either the fields still rely on expert models, or the progress of general foundation models lags significantly compared to those in popular areas, far from sufficient for transforming scientific research and leaving substantial gap between open-source models and closed-source models in these scientific domains. To mitigate this gap and explore a step further toward Artificial General Intelligence (AGI), we introduce Intern-S1, a specialized generalist equipped with general understanding and reasoning capabilities with expertise to analyze multiple science modal data. Intern-S1 is a multimodal Mixture-of-Experts (MoE) model with 28 billion activated parameters and 241 billion total parameters, continually pre-trained on 5T tokens, including over 2.5T tokens from scientific domains. In the post-training stage, Intern-S1 undergoes offline and then online reinforcement learning (RL) in InternBootCamp, where we propose Mixture-of-Rewards (MoR) to synergize the RL training on more than 1000 tasks simultaneously. Through integrated innovations in algorithms, data, and training systems, Intern-S1 achieved top-tier performance in online RL training. On comprehensive evaluation benchmarks, Intern-S1 demonstrates competitive performance on general reasoning tasks among open-source models and significantly outperforms open-source models in scientific domains, surpassing closed-source state-of-the-art models in professional tasks, such as molecular synthesis planning, reaction condition prediction, predicting thermodynamic stabilities for crystals. Our models are available at https://huggingface.co/internlm/Intern-S1.

Shuwei Chen, Xiang Li, Jian Dong et al.

Click-through rate (CTR) prediction plays a pivotal role in the success of recommendations. Inspired by the recent thriving of language models (LMs), a surge of works improve prediction by organizing user behavior data in a \textbf{textual} format and using LMs to understand user interest at a semantic level. While promising, these works have to truncate the textual data to reduce the quadratic computational overhead of self-attention in LMs. However, it has been studied that long user behavior data can significantly benefit CTR prediction. In addition, these works typically condense user diverse interests into a single feature vector, which hinders the expressive capability of the model. In this paper, we propose a \textbf{T}extual \textbf{B}ehavior-based \textbf{I}nterest Chunking \textbf{N}etwork (TBIN), which tackles the above limitations by combining an efficient locality-sensitive hashing algorithm and a shifted chunk-based self-attention. The resulting user diverse interests are dynamically activated, producing user interest representation towards the target item. Finally, the results of both offline and online experiments on real-world food recommendation platform demonstrate the effectiveness of TBIN.

Chao Wang, Yuqing Cai, Renzeng Duojie et al.

In this work, we propose a streaming speech recognition framework for Amdo Tibetan, built upon a hybrid CTC/Atten-tion architecture with a context-aware dynamic chunking mechanism. The proposed strategy adaptively adjusts chunk widths based on encoding states, enabling flexible receptive fields, cross-chunk information exchange, and robust adaptation to varying speaking rates, thereby alleviating the context truncation problem of fixed-chunk methods. To further capture the linguistic characteristics of Tibetan, we construct a lexicon grounded in its orthographic principles, providing linguistically motivated modeling units. During decoding, an external language model is integrated to enhance semantic consistency and improve recognition of long sentences. Experimental results show that the proposed framework achieves a word error rate (WER) of 6.23% on the test set, yielding a 48.15% relative improvement over the fixed-chunk baseline, while significantly reducing recognition latency and maintaining performance close to global decoding.

Hanwen Shi, Jin Zhang, Kunpeng Zhang

Matching patients effectively and efficiently for clinical trials is a significant challenge due to the complexity and variability of patient profiles and trial criteria. This paper introduces \textbf{Multi-Agent for Knowledge Augmentation and Reasoning (MAKAR)}, a novel multi-agent system that enhances patient-trial matching by integrating criterion augmentation with structured reasoning. MAKAR consistently improves performance by an average of 7\% across different datasets. Furthermore, it enables privacy-preserving deployment and maintains competitive performance when using smaller open-source models. Overall, MAKAR can contributes to more transparent, accurate, and privacy-conscious AI-driven patient matching.

Yongcun Song, Shangzhi Zeng, Jin Zhang et al.

Optimal control of obstacle problems arises in a wide range of applications and is computationally challenging due to its nonsmoothness, nonlinearity, and bilevel structure. Classical numerical approaches rely on mesh-based discretization and typically require solving a sequence of costly subproblems. In this work, we propose a single-loop bilevel deep learning method, which is mesh-free, scalable to high-dimensional and complex domains, and avoids repeated solution of discretized subproblems. The method employs constraint-embedding neural networks to approximate the state and control and preserves the bilevel structure. To train the neural networks efficiently, we propose a Single-Loop Stochastic First-Order Bilevel Algorithm (S2-FOBA), which eliminates nested optimization and does not rely on restrictive lower-level uniqueness assumptions. We analyze the convergence behavior of S2-FOBA under mild assumptions. Numerical experiments on benchmark examples, including distributed and obstacle control problems with regular and irregular obstacles on complex domains, demonstrate that the proposed method achieves satisfactory accuracy while reducing computational cost compared to classical numerical methods.

Jin Zhang

We propose the estimates of the discrete Green function for the stream- line diffusion finite element method (SDFEM) on Shishkin meshes.

Gaocheng Yu, Jin Zhang, Zhe Ma et al.

HDR is an important part of computational photography technology. In this paper, we propose a lightweight neural network called Efficient Attention-and-alignment-guided Progressive Network (EAPNet) for the challenge NTIRE 2022 HDR Track 1 and Track 2. We introduce a multi-dimensional lightweight encoding module to extract features. Besides, we propose Progressive Dilated U-shape Block (PDUB) that can be a progressive plug-and-play module for dynamically tuning MAccs and PSNR. Finally, we use fast and low-power feature-align module to deal with misalignment problem in place of the time-consuming Deformable Convolutional Network (DCN). The experiments show that our method achieves about 20 times compression on MAccs with better mu-PSNR and PSNR compared to the state-of-the-art method. We got the second place of both two tracks during the testing phase. Figure1. shows the visualized result of NTIRE 2022 HDR challenge.

Liu Qian, Yue Li, Ying Xie et al.

Catalyst design is crucial for materials synthesis, especially for complex reaction networks. Strategies like collaborative catalytic systems and multifunctional catalysts are effective but face challenges at the nanoscale. Carbon nanotube synthesis contains complicated nanoscale catalytic reactions, thus achieving high-density, high-quality semiconducting CNTs demands innovative catalyst design. In this work, we present a holistic framework integrating machine learning into traditional catalyst design for semiconducting CNT synthesis. It combines knowledge-based insights with data-driven techniques. Three key components, including open-access electronic structure databases for precise physicochemical descriptors, pre-trained natural language processing-based embedding model for higher-level abstractions, and physical - driven predictive models based on experiment data, are utilized. Through this framework, a new method for selective semiconducting CNT synthesis via catalyst - mediated electron injection, tuned by light during growth, is proposed. 54 candidate catalysts are screened, and three with high potential are identified. High-throughput experiments validate the predictions, with semiconducting selectivity exceeding 91% and the FeTiO3 catalyst reaching 98.6%. This approach not only addresses semiconducting CNT synthesis but also offers a generalizable methodology for global catalyst design and nanomaterials synthesis, advancing materials science in precise control.

Renyang Liu, Jinhong Zhang, Haoran Li et al.

Extensive studies have demonstrated that deep neural networks (DNNs) are vulnerable to adversarial attacks. Despite the significant progress in the attack success rate that has been made recently, the adversarial noise generated by most of the existing attack methods is still too conspicuous to the human eyes and proved to be easily detected by defense mechanisms. Resulting that these malicious examples cannot contribute to exploring the vulnerabilities of existing DNNs sufficiently. Thus, to better reveal the defects of DNNs and further help enhance their robustness under noise-limited situations, a new inconspicuous adversarial examples generation method is exactly needed to be proposed. To bridge this gap, we propose a novel Normalize Flow-based end-to-end attack framework, called AFLOW, to synthesize imperceptible adversarial examples under strict constraints. Specifically, rather than the noise-adding manner, AFLOW directly perturbs the hidden representation of the corresponding image to craft the desired adversarial examples. Compared with existing methods, extensive experiments on three benchmark datasets show that the adversarial examples built by AFLOW exhibit superiority in imperceptibility, image quality and attack capability. Even on robust models, AFLOW can still achieve higher attack results than previous methods.

Wenxuan Liu, Zixuan Li, Bai Long et al.

Deep Research (DR) requires LLM agents to autonomously perform multi-step information seeking, processing, and reasoning to generate comprehensive reports. In contrast to existing studies that mainly focus on unstructured web content, a more challenging DR task should additionally utilize structured knowledge to provide a solid data foundation, facilitate quantitative computation, and lead to in-depth analyses. In this paper, we refer to this novel task as Knowledgeable Deep Research (KDR), which requires DR agents to generate reports with both structured and unstructured knowledge. Furthermore, we propose the Hybrid Knowledge Analysis framework (HKA), a multi-agent architecture that reasons over both kinds of knowledge and integrates the texts, figures, and tables into coherent multimodal reports. The key design is the Structured Knowledge Analyzer, which utilizes both coding and vision-language models to produce figures, tables, and corresponding insights. To support systematic evaluation, we construct KDR-Bench, which covers 9 domains, includes 41 expert-level questions, and incorporates a large number of structured knowledge resources (e.g., 1,252 tables). We further annotate the main conclusions and key points for each question and propose three categories of evaluation metrics including general-purpose, knowledge-centric, and vision-enhanced ones. Experimental results demonstrate that HKA consistently outperforms most existing DR agents on general-purpose and knowledge-centric metrics, and even surpasses the Gemini DR agent on vision-enhanced metrics, highlighting its effectiveness in deep, structure-aware knowledge analysis. Finally, we hope this work can serve as a new foundation for structured knowledge analysis in DR agents and facilitate future multimodal DR studies.

Tianshu Chu, Dachuan Xu, Wei Yao et al.

While stochastic bilevel optimization methods have been extensively studied for addressing large-scale nested optimization problems in machine learning, it remains an open question whether the optimal complexity bounds for solving bilevel optimization are the same as those in single-level optimization. Our main result resolves this question: SPABA, an adaptation of the PAGE method for nonconvex optimization in (Li et al., 2021) to the bilevel setting, can achieve optimal sample complexity in both the finite-sum and expectation settings. We show the optimality of SPABA by proving that there is no gap in complexity analysis between stochastic bilevel and single-level optimization when implementing PAGE. Notably, as indicated by the results of (Dagréou et al., 2022), there might exist a gap in complexity analysis when implementing other stochastic gradient estimators, like SGD and SAGA. In addition to SPABA, we propose several other single-loop stochastic bilevel algorithms, that either match or improve the state-of-the-art sample complexity results, leveraging our convergence rate and complexity analysis. Numerical experiments demonstrate the superior practical performance of the proposed methods.

Jin Zhang

We propose estimates of the discrete Green function for the streamline diffusion finite element method (SDFEM) on Shishkin triangular meshes.

Kimi Team, Angang Du, Bofei Gao et al. · pku, tsinghua

Language model pretraining with next token prediction has proved effective for scaling compute but is limited to the amount of available training data. Scaling reinforcement learning (RL) unlocks a new axis for the continued improvement of artificial intelligence, with the promise that large language models (LLMs) can scale their training data by learning to explore with rewards. However, prior published work has not produced competitive results. In light of this, we report on the training practice of Kimi k1.5, our latest multi-modal LLM trained with RL, including its RL training techniques, multi-modal data recipes, and infrastructure optimization. Long context scaling and improved policy optimization methods are key ingredients of our approach, which establishes a simplistic, effective RL framework without relying on more complex techniques such as Monte Carlo tree search, value functions, and process reward models. Notably, our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities -- e.g., 77.5 on AIME, 96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching OpenAI's o1. Moreover, we present effective long2short methods that use long-CoT techniques to improve short-CoT models, yielding state-of-the-art short-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on LiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and Claude Sonnet 3.5 by a large margin (up to +550%).

Zhuoyuan Ma, Qi Zhao, Jin Zhang et al.

Aerial simultaneous transmitting and reflecting reconfigurable intelligent surfaces (STAR-RIS) enables full-space coverage in dynamic wireless networks. However, most existing works assume fixed user grouping, overlooking the fact that STAR-RIS deployment inherently determines whether users are served via transmission or reflection. To address this, we propose a joint deployment and beamforming framework, where an aerial STAR-RIS dynamically adjusts its location and orientation to adaptively control user grouping and enhance hybrid beamforming. We formulate a Markov decision process (MDP) capturing the coupling among deployment, grouping, and signal design. To solve the resulting non-convex and time-varying problem, we develop a PPO-based reinforcement learning algorithm that adaptively balances user grouping and beamforming resources through online policy learning. Simulation results show 57.1\% and 285\% sum-rate gains over fixed-deployment and RIS-free baselines, respectively, demonstrating the benefit of user-grouping-aware control in STAR-RIS-aided systems.

Ruisong Gao, Min Yang, Jin Zhang

We propose a new class of physics-informed neural networks, called Physics-Informed Generator-Encoder Adversarial Networks, to effectively address the challenges posed by forward, inverse, and mixed problems in stochastic differential equations. In these scenarios, while the governing equations are known, the available data consist of only a limited set of snapshots for system parameters. Our model consists of two key components: the generator and the encoder, both updated alternately by gradient descent. In contrast to previous approaches of directly matching the approximated solutions with real snapshots, we employ an indirect matching that operates within the lower-dimensional latent feature space. This method circumvents challenges associated with high-dimensional inputs and complex data distributions, while yielding more accurate solutions compared to existing neural network solvers. In addition, the approach also mitigates the training instability issues encountered in previous adversarial frameworks in an efficient manner. Numerical results provide compelling evidence of the effectiveness of the proposed method in solving different types of stochastic differential equations.

Liyuan Huang, Jiawei He, Wutao Shen et al.

This paper describes our system to SemEval-2026 Task 3 Track A Subtask 1 on Dimensional Aspect Sentiment Regression (DimASR). We propose a lightweight and resource-efficient system built entirely on multilingual pre-trained encoders, without relying on LLMs or external corpora. We adopt joint multilingual and multi-domain training to facilitate cross-lingual transfer and alleviate data sparsity, introduce a bounded regression transformation that improves training stability while constraining predictions within the valid range, and employ an adaptive ensemble strategy via subset search to reduce prediction variance. Experimental results demonstrate that our system achieves strong and consistent performance, ranking 1st on zho-res, 2nd on zho-lap, and 3rd on jpn-hot, with all remaining datasets placed within the top half of participating teams.

Yiqin Yang, Hao Hu, Yihuan Mao et al.

Preference-based reinforcement learning (PbRL) can help avoid sophisticated reward designs and align better with human intentions, showing great promise in various real-world applications. However, obtaining human feedback for preferences can be expensive and time-consuming, which forms a strong barrier for PbRL. In this work, we address the problem of low query efficiency in offline PbRL, pinpointing two primary reasons: inefficient exploration and overoptimization of learned reward functions. In response to these challenges, we propose a novel algorithm, \textbf{O}ffline \textbf{P}b\textbf{R}L via \textbf{I}n-\textbf{D}ataset \textbf{E}xploration (OPRIDE), designed to enhance the query efficiency of offline PbRL. OPRIDE consists of two key features: a principled exploration strategy that maximizes the informativeness of the queries and a discount scheduling mechanism aimed at mitigating overoptimization of the learned reward functions. Through empirical evaluations, we demonstrate that OPRIDE significantly outperforms prior methods, achieving strong performance with notably fewer queries. Moreover, we provide theoretical guarantees of the algorithm's efficiency. Experimental results across various locomotion, manipulation, and navigation tasks underscore the efficacy and versatility of our approach.

Junjie Gao, Runze Liu, Yingzhe Peng et al.

Document quality assessment is critical for a wide range of applications including document digitization, OCR, and archival. However, existing approaches often struggle to provide accurate and robust quality scores, limiting their applicability in practical scenarios. With the rapid progress in Multi-modal Large Language Models (MLLMs), recent MLLM-based methods have achieved remarkable performance in image quality assessment. In this work, we extend this success to the document domain by adapting DeQA-Score, a state-of-the-art MLLM-based image quality scorer, for document quality assessment. We propose DeQA-Doc, a framework that leverages the visual language capabilities of MLLMs and a soft label strategy to regress continuous document quality scores. To adapt DeQA-Score to DeQA-Doc, we adopt two complementary solutions to construct soft labels without the variance information. Also, we relax the resolution constrains to support the large resolution of document images. Finally, we introduce ensemble methods to further enhance the performance. Extensive experiments demonstrate that DeQA-Doc significantly outperforms existing baselines, offering accurate and generalizable document quality assessment across diverse degradation types. Codes and model weights are available in https://github.com/Junjie-Gao19/DeQA-Doc.

Risheng Liu, Pan Mu, Xiaoming Yuan et al.

In recent years, a variety of gradient-based methods have been developed to solve Bi-Level Optimization (BLO) problems in machine learning and computer vision areas. However, the theoretical correctness and practical effectiveness of these existing approaches always rely on some restrictive conditions (e.g., Lower-Level Singleton, LLS), which could hardly be satisfied in real-world applications. Moreover, previous literature only proves theoretical results based on their specific iteration strategies, thus lack a general recipe to uniformly analyze the convergence behaviors of different gradient-based BLOs. In this work, we formulate BLOs from an optimistic bi-level viewpoint and establish a new gradient-based algorithmic framework, named Bi-level Descent Aggregation (BDA), to partially address the above issues. Specifically, BDA provides a modularized structure to hierarchically aggregate both the upper- and lower-level subproblems to generate our bi-level iterative dynamics. Theoretically, we establish a general convergence analysis template and derive a new proof recipe to investigate the essential theoretical properties of gradient-based BLO methods. Furthermore, this work systematically explores the convergence behavior of BDA in different optimization scenarios, i.e., considering various solution qualities (i.e., global/local/stationary solution) returned from solving approximation subproblems. Extensive experiments justify our theoretical results and demonstrate the superiority of the proposed algorithm for hyper-parameter optimization and meta-learning tasks. Source code is available at https://github.com/vis-opt-group/BDA.