Xin Huang

Jihao Liu, Xin Huang, Jinliang Zheng et al. · tsinghua

In this paper, we propose Mixed and Masked AutoEncoder (MixMAE), a simple but efficient pretraining method that is applicable to various hierarchical Vision Transformers. Existing masked image modeling (MIM) methods for hierarchical Vision Transformers replace a random subset of input tokens with a special [MASK] symbol and aim at reconstructing original image tokens from the corrupted image. However, we find that using the [MASK] symbol greatly slows down the training and causes pretraining-finetuning inconsistency, due to the large masking ratio (e.g., 60% in SimMIM). On the other hand, MAE does not introduce [MASK] tokens at its encoder at all but is not applicable for hierarchical Vision Transformers. To solve the issue and accelerate the pretraining of hierarchical models, we replace the masked tokens of one image with visible tokens of another image, i.e., creating a mixed image. We then conduct dual reconstruction to reconstruct the two original images from the mixed input, which significantly improves efficiency. While MixMAE can be applied to various hierarchical Transformers, this paper explores using Swin Transformer with a large window size and scales up to huge model size (to reach 600M parameters). Empirical results demonstrate that MixMAE can learn high-quality visual representations efficiently. Notably, MixMAE with Swin-B/W14 achieves 85.1% top-1 accuracy on ImageNet-1K by pretraining for 600 epochs. Besides, its transfer performances on the other 6 datasets show that MixMAE has better FLOPs / performance tradeoff than previous popular MIM methods. Code is available at https://github.com/Sense-X/MixMIM.

Yifan Sui, Xin Huang, Hongbing Li et al.

The rapid development of GUI foundation models and mobile GUI agents has spurred numerous evaluation benchmarks, yet most rely on simulated environments or open-source applications, leaving real-world closed-source applications largely unevaluated. The core difficulty is that closed-source applications do not expose internal states, making traditional automatic verification inapplicable. To bridge this gap, we introduce AndroidDaily, a large-scale benchmark comprising 350 realistic daily-use tasks across 94 high-frequency Android applications spanning transportation, shopping, local services, entertainment, content creation, social media, and everyday utilities. To enable automatic and verifiable assessment in these opaque environments, we propose Guideline-grounded Reviewer for Automatic Diagnostic Evaluation (GRADE), a process-aware evaluator built on a three-tiered system of observable external guidelines: operational obligations, output quality, and negative constraints. GRADE tracks the agent's visual trajectory against these criteria and produces step-level diagnostic judgments, turning long-horizon, open-ended mobile interactions into verifiable evaluation without relying on hidden internal states. Experiments show that GRADE achieves 87.37\% agreement with human evaluators. The strongest model reaches a 62.0\% success rate on AndroidDaily, highlighting a substantial gap between current reasoning capabilities and practical execution in realistic mobile workflows.

Jihao Liu, Xin Huang, Guanglu Song et al.

Recently, transformer and multi-layer perceptron (MLP) architectures have achieved impressive results on various vision tasks. However, how to effectively combine those operators to form high-performance hybrid visual architectures still remains a challenge. In this work, we study the learnable combination of convolution, transformer, and MLP by proposing a novel unified architecture search approach. Our approach contains two key designs to achieve the search for high-performance networks. First, we model the very different searchable operators in a unified form, and thus enable the operators to be characterized with the same set of configuration parameters. In this way, the overall search space size is significantly reduced, and the total search cost becomes affordable. Second, we propose context-aware downsampling modules (DSMs) to mitigate the gap between the different types of operators. Our proposed DSMs are able to better adapt features from different types of operators, which is important for identifying high-performance hybrid architectures. Finally, we integrate configurable operators and DSMs into a unified search space and search with a Reinforcement Learning-based search algorithm to fully explore the optimal combination of the operators. To this end, we search a baseline network and scale it up to obtain a family of models, named UniNets, which achieve much better accuracy and efficiency than previous ConvNets and Transformers. In particular, our UniNet-B5 achieves 84.9% top-1 accuracy on ImageNet, outperforming EfficientNet-B7 and BoTNet-T7 with 44% and 55% fewer FLOPs respectively. By pretraining on the ImageNet-21K, our UniNet-B6 achieves 87.4%, outperforming Swin-L with 51% fewer FLOPs and 41% fewer parameters. Code is available at https://github.com/Sense-X/UniNet.

Bin Wang, Zhengyuan Liu, Xin Huang et al.

We present SeaEval, a benchmark for multilingual foundation models. In addition to characterizing how these models understand and reason with natural language, we also investigate how well they comprehend cultural practices, nuances, and values. Alongside standard accuracy metrics, we investigate the brittleness of foundation models in the dimensions of semantics and multilinguality. Our analyses span both open-sourced and closed models, leading to empirical results across classic NLP tasks, reasoning, and cultural comprehension. Key findings indicate (1) Most models exhibit varied behavior when given paraphrased instructions. (2) Many models still suffer from exposure bias (e.g., positional bias, majority label bias). (3) For questions rooted in factual, scientific, and commonsense knowledge, consistent responses are expected across multilingual queries that are semantically equivalent. Yet, most models surprisingly demonstrate inconsistent performance on these queries. (4) Multilingually-trained models have not attained "balanced multilingual" capabilities. Our endeavors underscore the need for more generalizable semantic representations and enhanced multilingual contextualization. SeaEval can serve as a launchpad for more thorough investigations and evaluations for multilingual and multicultural scenarios.

Hao Zhou, Zhijun Wang, Shujian Huang et al.

Large Language Models (LLMs) are often English-centric due to the disproportionate distribution of languages in their pre-training data. Enhancing non-English language capabilities through post-pretraining often results in catastrophic forgetting of the ability of original languages. Previous methods either achieve good expansion with severe forgetting or slight forgetting with poor expansion, indicating the challenge of balancing language expansion while preventing forgetting. In this paper, we propose a method called MoE-LPR (Mixture-of-Experts with Language Priors Routing) to alleviate this problem. MoE-LPR employs a two-stage training approach to enhance the multilingual capability. First, the model is post-pretrained into a Mixture-of-Experts (MoE) architecture by upcycling, where all the original parameters are frozen and new experts are added. In this stage, we focus improving the ability on expanded languages, without using any original language data. Then, the model reviews the knowledge of the original languages with replay data amounting to less than 1% of post-pretraining, where we incorporate language priors routing to better recover the abilities of the original languages. Evaluations on multiple benchmarks show that MoE-LPR outperforms other post-pretraining methods. Freezing original parameters preserves original language knowledge while adding new experts preserves the learning ability. Reviewing with LPR enables effective utilization of multilingual knowledge within the parameters. Additionally, the MoE architecture maintains the same inference overhead while increasing total model parameters. Extensive experiments demonstrate MoE-LPR's effectiveness in improving expanded languages and preserving original language proficiency with superior scalability. Code and scripts are freely available at https://github.com/zjwang21/MoE-LPR.git.

Xin Huang, Aku Kammonen, Anamika Pandey et al.

The machine learning random Fourier feature method for data in high dimension is computationally and theoretically attractive since the optimization is based on a convex standard least squares problem and independent sampling of Fourier frequencies. The challenge is to sample the Fourier frequencies well. This work proves convergence of a data adaptive method based on resampling the frequencies asymptotically optimally, as the number of nodes and amount of data tend to infinity. Numerical results based on resampling and adaptive random walk steps together with approximations of the least squares problem by conjugate gradient iterations confirm the analysis for regression and classification problems.

Xin Huang, Antoni B. Chan

Large Language Models (LLMs) are increasingly evaluated with input attribution methods, yet comparing such explanations remains challenging. Existing soft-perturbation faithfulness metrics, such as Soft-NC and Soft-NS, can conflate attribution quality with the number of words retained during perturbation: attribution methods with larger average scores may keep more words and therefore obtain inflated scores. To address this issue, we propose $π$-Soft-NC and $π$-Soft-NS, an evaluation framework that compares attribution methods under the same expected retaining probability, thus controlling the number of retained words. We further introduce Grad-ELLM, a gradient-based attribution method tailored to autoregressive decoder-only LLMs, which combines gradient-derived channel importance with attention-derived token importance at each decoding step. Experiments on classification and open-generation tasks with Llama and Mistral show that Grad-ELLM achieves strong comprehensiveness-oriented faithfulness under $π$-Soft-NC, while there is no dominant method under $π$-Soft-NS. Our evaluation metric serves as a rigorous framework to compare XAI methods for LLMs, which will support progress in the field.

Ailin Huang, Chengyuan Yao, Chunrui Han et al.

We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-VL-10B is realized through two strategic shifts: first, a unified, fully unfrozen pre-training strategy on 1.2T multimodal tokens that integrates a language-aligned Perception Encoder with a Qwen3-8B decoder to establish intrinsic vision-language synergy; and second, a scaled post-training pipeline featuring over 1k iterations of reinforcement learning. Crucially, we implement Parallel Coordinated Reasoning (PaCoRe) to scale test-time compute, allocating resources to scalable perceptual reasoning that explores and synthesizes diverse visual hypotheses. Consequently, despite its compact 10B footprint, STEP3-VL-10B rivals or surpasses models 10$\times$-20$\times$ larger (e.g., GLM-4.6V-106B, Qwen3-VL-235B) and top-tier proprietary flagships like Gemini 2.5 Pro and Seed-1.5-VL. Delivering best-in-class performance, it records 92.2% on MMBench and 80.11% on MMMU, while excelling in complex reasoning with 94.43% on AIME2025 and 75.95% on MathVision. We release the full model suite to provide the community with a powerful, efficient, and reproducible baseline.

Yuxi Sun, Wenbo Shang, Wei Gao et al.

In RAG-based fact-checking, LLMs are increasingly used as verifiers to check given claims against retrieved evidence. Their parametric knowledge can induce pre-evidence tendencies that may conflict with the retrieved context, yet existing evaluation frameworks do not characterize such prior-context discrepancy or measure how verifiers arbitrate between parametric and contextual signals. We introduce \textsc{PAVE} (\emph{Prior-Aware Verifier Evaluation}), a diagnostic testbed that stratifies an LLM verifier into four epistemic states based on the correctness and confidence of its pre-evidence prior and evaluates its arbitration behavior on this new benchmark, i.e., whether it persists in correct prior under misleading evidence, and whether it corrects wrong prior when accurate evidence is provided. Experiments across seven LLMs reveal unreliable and highly model-dependent prior-context arbitration, highlighting the importance of verifier selection for real-world RAG-based fact-checking applications. Based on these findings, we propose a lightweight JSD-based test-time arbitration method that improves factual reliability without modifying the underlying model, achieving competitive performance across diverse LLM families.

Xin Huang, Ruizhi Shao, Qi Zhang et al.

Recent text-to-3D methods employing diffusion models have made significant advancements in 3D human generation. However, these approaches face challenges due to the limitations of text-to-image diffusion models, which lack an understanding of 3D structures. Consequently, these methods struggle to achieve high-quality human generation, resulting in smooth geometry and cartoon-like appearances. In this paper, we propose HumanNorm, a novel approach for high-quality and realistic 3D human generation. The main idea is to enhance the model's 2D perception of 3D geometry by learning a normal-adapted diffusion model and a normal-aligned diffusion model. The normal-adapted diffusion model can generate high-fidelity normal maps corresponding to user prompts with view-dependent and body-aware text. The normal-aligned diffusion model learns to generate color images aligned with the normal maps, thereby transforming physical geometry details into realistic appearance. Leveraging the proposed normal diffusion model, we devise a progressive geometry generation strategy and a multi-step Score Distillation Sampling (SDS) loss to enhance the performance of 3D human generation. Comprehensive experiments substantiate HumanNorm's ability to generate 3D humans with intricate geometry and realistic appearances. HumanNorm outperforms existing text-to-3D methods in both geometry and texture quality. The project page of HumanNorm is https://humannorm.github.io/.

Wenwen Zhuang, Xin Huang, Xiantao Zhang et al.

Multimodal Large Language Models (MLLMs) excel in solving text-based mathematical problems, but they struggle with mathematical diagrams since they are primarily trained on natural scene images. For humans, visual aids generally enhance problem-solving, but MLLMs perform worse as information shifts from textual to visual modality. This decline is mainly due to their shortcomings in aligning images and text. To tackle aforementioned challenges, we propose Math-PUMA, a methodology focused on Progressive Upward Multimodal Alignment. This approach is designed to improve the mathematical reasoning skills of MLLMs through a three-stage training process, with the second stage being the critical alignment stage. We first enhance the language model's mathematical reasoning capabilities with extensive set of textual mathematical problems. We then construct a multimodal dataset with varying degrees of textual and visual information, creating data pairs by presenting each problem in at least two forms. By leveraging the Kullback-Leibler (KL) divergence of next-token prediction distributions to align visual and textual modalities, consistent problem-solving abilities are ensured. Finally, we utilize multimodal instruction tuning for MLLMs with high-quality multimodal data. Experimental results on multiple mathematical reasoning benchmarks demonstrate that the MLLMs trained with Math-PUMA surpass most open-source MLLMs. Our approach effectively narrows the performance gap for problems presented in different modalities. The code and data are available at: \url{https://github.com/wwzhuang01/Math-PUMA}.

Yipeng Ouyang, Xin Huang, Bingjie Liu et al.

LLM agents are rapidly evolving from coding assistants into autonomous software engineering systems. However, existing evaluation methodologies remain largely centered on static, isolated, and short-horizon benchmarks that fail to capture the dynamic complexity of real-world production workflows. As a result, benchmark performance may poorly reflect practical capability under realistic runtime environments involving long execution chains, tool interactions, dependency management, and iterative feedback loops. We thus present RAMP, a production-grounded infrastructure for assessing long-horizon software engineering agents. Built upon the YatCC integrated platform, RAMP provides a unified runtime assessment architecture through standardized orchestration and execution interfaces. RAMP introduces realistic compiler-construction workloads with serial dependencies and complex toolchain interactions, together with a staged recovery mechanism for analyzing execution behavior under partial workflow failure. The framework further incorporates utility-oriented multi-dimensional metrics that jointly evaluate outcome quality and process efficiency. We conduct runtime assessments across 15 mainstream models and observe substantial capability degradation that remains largely invisible to conventional isolated benchmarks. Task completion rates progressively collapse across serial workflows, dropping from 100% in the initial stage to only 20% in the final stage, while none of the evaluated models successfully completes the entire pipeline. Runtime analysis reveals systematic failure propagation and significant resource inefficiencies, with computational costs differing by up to three orders of magnitude among comparable models. These findings suggest RAMP advances agentic model evaluation toward continuous, runtime-observable, and production-grounded assessment.

Qiao Sun, Xin Huang, Brian C. Williams et al.

Motion prediction is crucial in enabling safe motion planning for autonomous vehicles in interactive scenarios. It allows the planner to identify potential conflicts with other traffic agents and generate safe plans. Existing motion predictors often focus on reducing prediction errors, yet it remains an open question on how well they help identify the conflicts for the planner. In this paper, we evaluate state-of-the-art predictors through novel conflict-related metrics, such as the success rate of identifying conflicts. Surprisingly, the predictors suffer from a low success rate and thus lead to a large percentage of collisions when we test the prediction-planning system in an interactive simulator. To fill the gap, we propose a simple but effective alternative that combines a physics-based trajectory generator and a learning-based relation predictor to identify conflicts and infer conflict relations. We demonstrate that our predictor, P4P, achieves superior performance over existing learning-based predictors in realistic interactive driving scenarios from Waymo Open Motion Dataset.

Xin Huang, Ashish Khetan, Rene Bidart et al.

Transformer-based language models such as BERT have achieved the state-of-the-art performance on various NLP tasks, but are computationally prohibitive. A recent line of works use various heuristics to successively shorten sequence length while transforming tokens through encoders, in tasks such as classification and ranking that require a single token embedding for prediction. We present a novel solution to this problem, called Pyramid-BERT where we replace previously used heuristics with a {\em core-set} based token selection method justified by theoretical results. The core-set based token selection technique allows us to avoid expensive pre-training, gives a space-efficient fine tuning, and thus makes it suitable to handle longer sequence lengths. We provide extensive experiments establishing advantages of pyramid BERT over several baselines and existing works on the GLUE benchmarks and Long Range Arena datasets.

Xin Huang, Qi Zhang, Ying Feng et al.

Representing visual signals with implicit coordinate-based neural networks, as an effective replacement of the traditional discrete signal representation, has gained considerable popularity in computer vision and graphics. In contrast to existing implicit neural representations which focus on modelling the scene only, this paper proposes a novel implicit camera model which represents the physical imaging process of a camera as a deep neural network. We demonstrate the power of this new implicit camera model on two inverse imaging tasks: i) generating all-in-focus photos, and ii) HDR imaging. Specifically, we devise an implicit blur generator and an implicit tone mapper to model the aperture and exposure of the camera's imaging process, respectively. Our implicit camera model is jointly learned together with implicit scene models under multi-focus stack and multi-exposure bracket supervision. We have demonstrated the effectiveness of our new model on a large number of test images and videos, producing accurate and visually appealing all-in-focus and high dynamic range images. In principle, our new implicit neural camera model has the potential to benefit a wide array of other inverse imaging tasks.

Majid Khonji, Rashid Alyassi, Wolfgang Merkt et al.

In transportation networks, where traffic lights have traditionally been used for vehicle coordination, intersections act as natural bottlenecks. A formidable challenge for existing automated intersections lies in detecting and reasoning about uncertainty from the operating environment and human-driven vehicles. In this paper, we propose a risk-aware intelligent intersection system for autonomous vehicles (AVs) as well as human-driven vehicles (HVs). We cast the problem as a novel class of Multi-agent Chance-Constrained Stochastic Shortest Path (MCC-SSP) problems and devise an exact Integer Linear Programming (ILP) formulation that is scalable in the number of agents' interaction points (e.g., potential collision points at the intersection). In particular, when the number of agents within an interaction point is small, which is often the case in intersections, the ILP has a polynomial number of variables and constraints. To further improve the running time performance, we show that the collision risk computation can be performed offline. Additionally, a trajectory optimization workflow is provided to generate risk-aware trajectories for any given intersection. The proposed framework is implemented in CARLA simulator and evaluated under a fully autonomous intersection with AVs only as well as in a hybrid setup with a signalized intersection for HVs and an intelligent scheme for AVs. As verified via simulations, the featured approach improves intersection's efficiency by up to $200\%$ while also conforming to the specified tunable risk threshold.

Xiuling Wang, Xin Huang, Guibo Luo et al.

Graph generative diffusion models have recently emerged as a powerful paradigm for generating complex graph structures, effectively capturing intricate dependencies and relationships within graph data. However, the privacy risks associated with these models remain largely unexplored. In this paper, we investigate information leakage in such models through three types of black-box inference attacks. First, we design a graph reconstruction attack, which can reconstruct graphs structurally similar to those training graphs from the generated graphs. Second, we propose a property inference attack to infer the properties of the training graphs, such as the average graph density and the distribution of densities, from the generated graphs. Third, we develop two membership inference attacks to determine whether a given graph is present in the training set. Extensive experiments on three different types of graph generative diffusion models and six real-world graphs demonstrate the effectiveness of these attacks, significantly outperforming the baseline approaches. Finally, we propose two defense mechanisms that mitigate these inference attacks and achieve a better trade-off between defense strength and target model utility than existing methods. Our code is available at https://zenodo.org/records/17946102.

Cheng Lyu, Jiake Xie, Bo Xu et al.

Performance of trimap-free image matting methods is limited when trying to decouple the deterministic and undetermined regions, especially in the scenes where foregrounds are semantically ambiguous, chromaless, or high transmittance. In this paper, we propose a novel framework named Privileged Prior Information Distillation for Image Matting (PPID-IM) that can effectively transfer privileged prior environment-aware information to improve the performance of students in solving hard foregrounds. The prior information of trimap regulates only the teacher model during the training stage, while not being fed into the student network during actual inference. In order to achieve effective privileged cross-modality (i.e. trimap and RGB) information distillation, we introduce a Cross-Level Semantic Distillation (CLSD) module that reinforces the trimap-free students with more knowledgeable semantic representations and environment-aware information. We also propose an Attention-Guided Local Distillation module that efficiently transfers privileged local attributes from the trimap-based teacher to trimap-free students for the guidance of local-region optimization. Extensive experiments demonstrate the effectiveness and superiority of our PPID framework on the task of image matting. In addition, our trimap-free IndexNet-PPID surpasses the other competing state-of-the-art methods by a large margin, especially in scenarios with chromaless, weak texture, or irregular objects.

Xin Huang, Qi Zhang, Ying Feng et al.

We propose LIRF (Local Implicit Ray Function), a generalizable neural rendering approach for novel view rendering. Current generalizable neural radiance fields (NeRF) methods sample a scene with a single ray per pixel and may therefore render blurred or aliased views when the input views and rendered views capture scene content with different resolutions. To solve this problem, we propose LIRF to aggregate the information from conical frustums to construct a ray. Given 3D positions within conical frustums, LIRF takes 3D coordinates and the features of conical frustums as inputs and predicts a local volumetric radiance field. Since the coordinates are continuous, LIRF renders high-quality novel views at a continuously-valued scale via volume rendering. Besides, we predict the visible weights for each input view via transformer-based feature matching to improve the performance in occluded areas. Experimental results on real-world scenes validate that our method outperforms state-of-the-art methods on novel view rendering of unseen scenes at arbitrary scales.

Xiuling Wang, Xin Huang, Haibo Hu et al.

Graph Neural Networks (GNNs) unlock new ways of learning from graph-structured data, proving highly effective in capturing complex relationships and patterns. Federated GNNs (FGNNs) have emerged as a prominent distributed learning paradigm for training GNNs over decentralized data. However, FGNNs face two significant challenges: high communication overhead from multiple rounds of parameter exchanges and non-IID data characteristics across clients. To address these issues, we introduce CeFGC, a novel FGNN paradigm that facilitates efficient GNN training over non-IID data by limiting communication between the server and clients to three rounds only. The core idea of CeFGC is to leverage generative diffusion models to minimize direct client-server communication. Each client trains a generative diffusion model that captures its local graph distribution and shares this model with the server, which then redistributes it back to all clients. Using these generative models, clients generate synthetic graphs combined with their local graphs to train local GNN models. Finally, clients upload their model weights to the server for aggregation into a global GNN model. We theoretically analyze the I/O complexity of communication volume to show that CeFGC reduces to a constant of three communication rounds only. Extensive experiments on several real graph datasets demonstrate the effectiveness and efficiency of CeFGC against state-of-the-art competitors, reflecting our superior performance on non-IID graphs by aligning local and global model objectives and enriching the training set with diverse graphs.

Xin Huang, Xiaoyu Tian, Junru Gu et al.

Predicting future behaviors of road agents is a key task in autonomous driving. While existing models have demonstrated great success in predicting marginal agent future behaviors, it remains a challenge to efficiently predict consistent joint behaviors of multiple agents. Recently, the occupancy flow fields representation was proposed to represent joint future states of road agents through a combination of occupancy grid and flow, which supports efficient and consistent joint predictions. In this work, we propose a novel occupancy flow fields predictor to produce accurate occupancy and flow predictions, by combining the power of an image encoder that learns features from a rasterized traffic image and a vector encoder that captures information of continuous agent trajectories and map states. The two encoded features are fused by multiple attention modules before generating final predictions. Our simple but effective model ranks 3rd place on the Waymo Open Dataset Occupancy and Flow Prediction Challenge, and achieves the best performance in the occluded occupancy and flow prediction task.

Xin Huang, Junjie Liang, Qingshan Hou et al.

Medical image synthesis is crucial for alleviating data scarcity and privacy constraints. However, fine-tuning general text-to-image (T2I) models remains challenging, mainly due to the significant modality gap between complex visual details and abstract clinical text. In addition, semantic entanglement persists, where coarse-grained text embeddings blur the boundary between anatomical structures and imaging styles, thus weakening controllability during generation. To address this, we propose a Visually-Guided Text Disentanglement framework. We introduce a cross-modal latent alignment mechanism that leverages visual priors to explicitly disentangle unstructured text into independent semantic representations. Subsequently, a Hybrid Feature Fusion Module (HFFM) injects these features into a Diffusion Transformer (DiT) through separated channels, enabling fine-grained structural control. Experimental results in three datasets demonstrate that our method outperforms existing approaches in terms of generation quality and significantly improves performance on downstream classification tasks. The source code is available at https://github.com/hx111/VG-MedGen.

Ailin Huang, Boyong Wu, Bruce Wang et al.

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

Qi Zhang, Xin Huang, Ying Feng et al.

This paper studies the problem of view synthesis with certain amount of rotations from a pair of images, what we called stereo unstructured magnification. While the multi-plane image representation is well suited for view synthesis with depth invariant, how to generalize it to unstructured views remains a significant challenge. This is primarily due to the depth-dependency caused by camera frontal parallel representation. Here we propose a novel multiple homography image (MHI) representation, comprising of a set of scene planes with fixed normals and distances. A two-stage network is developed for novel view synthesis. Stage-1 is an MHI reconstruction module that predicts the MHIs and composites layered multi-normal images along the normal direction. Stage-2 is a normal-blending module to find blending weights. We also derive an angle-based cost to guide the blending of multi-normal images by exploiting per-normal geometry. Compared with the state-of-the-art methods, our method achieves superior performance for view synthesis qualitatively and quantitatively, especially for cases when the cameras undergo rotations.

Yang Fu, Yuliang Zou, Hao Xiang et al.

In autonomous driving, mapping is critical for motion planning but remains an under-utilized resource for perception tasks such as 3D object detection. Maps can provide robust structural priors of the static environment, helping resolve ambiguities and correct for sensor data sparsity or noise, especially for distant objects or under adverse weather conditions. However, conventional High-Definition (HD) maps are resource-intensive to obtain and maintain, which presents a challenge for efficient, large-scale deployment. In this paper, we propose a scalable solution to systematically leverage mapping to improve 3D detection by overcoming two primary challenges. First, we introduce a pipeline to automatically build dense mapping priors from aggregated sensor data, eliminating the need for human labeling. Second, we design a novel Mapping Priors Augmented 3D Detection (MPA3D) framework to effectively integrate mapping priors with different sensor modalities. Extensive experiments on the Waymo Open Dataset demonstrate that our approach achieves new state-of-the-art results, proving the effectiveness of scalable reconstructed scene priors for enhancing 3D detection.

Ziliang Chen, Xin Huang, Quanlong Guan et al.

The vision community is undergoing the unprecedented progress with the emergence of Vision-Language Pretraining Models (VLMs). Prompt learning plays as the holy grail of accessing VLMs since it enables their fast adaptation to downstream tasks with limited resources. Whereas existing researches milling around single-prompt paradigms, rarely investigate the technical potential behind their multi-prompt learning counterparts. This paper aims to provide a principled retrospect for vision-language multi-prompt learning. We extend the recent constant modality gap phenomenon to learnable prompts and then, justify the superiority of vision-language transfer with multi-prompt augmentation, empirically and theoretically. In terms of this observation, we propose an Energy-based Multi-prompt Learning (EMPL) to generate multiple prompt embeddings by drawing instances from an energy-based distribution, which is implicitly defined by VLMs. So our EMPL is not only parameter-efficient but also rigorously lead to the balance between in-domain and out-of-domain open-vocabulary generalization. Comprehensive experiments have been conducted to justify our claims and the excellence of EMPL.

Haolong Yan, Yeqing Shen, Xin Huang et al.

With the rapid development of Large Vision Language Models, the focus of Graphical User Interface (GUI) agent tasks shifts from single-screen tasks to complex screen navigation challenges. However, real-world GUI environments, such as PC software and mobile Apps, are often complex and proprietary, making it difficult to obtain the comprehensive environment information needed for agent training and evaluation. This limitation hinders systematic investigation and benchmarking of agent navigation capabilities. To address this limitation, we introduce GUI Exploration Lab, a simulation environment engine for GUI agent navigation research that enables flexible definition and composition of screens, icons, and navigation graphs, while providing full access to environment information for comprehensive agent training and evaluation. Through extensive experiments, we find that supervised fine-tuning enables effective memorization of fundamental knowledge, serving as a crucial foundation for subsequent training. Building on this, single-turn reinforcement learning further enhances generalization to unseen scenarios. Finally, multi-turn reinforcement learning encourages the development of exploration strategies through interactive trial and error, leading to further improvements in screen navigation performance. We validate our methods on both static and interactive benchmarks, demonstrating that our findings generalize effectively to real-world scenarios. These findings demonstrate the advantages of reinforcement learning approaches in GUI navigation and offer practical guidance for building more capable and generalizable GUI agents.

Yingwei Li, Xin Huang, Yang Liu et al.

Model scaling has demonstrated remarkable success through large-scale training on diverse datasets. It remains an open question whether the same paradigm would apply to autonomous driving perception systems due to unique challenges, such as fusing heterogeneous sensor data and the need for sophisticated 3D spatial understanding. To bridge this gap, we present a comprehensive study on systematically analyzing the impact of scale on these systems. We develop our STELLAR model based on Sparse Window Transformer, by extending the input modalities to include LiDAR, radar, camera, and map prior. We train the model on a large-scale dataset of 50 million driving examples with up to 500 million parameters. Our large-scale experiments reveal empirical scaling trends that connect model performance to model size, data, and compute. The resulting model establishes a new state-of-the-art on the Waymo Open Dataset challenge, outperforming prior arts by a large margin. Our work demonstrates that large-scale training is a highly promising path for advancing the capabilities of perception models for autonomous driving.

Mingyue Cheng, Hao Zhang, Jiqian Yang et al.

Large language model evaluation plays a pivotal role in the enhancement of its capacity. Previously, numerous methods for evaluating large language models have been proposed in this area. Despite their effectiveness, these existing works mainly focus on assessing objective questions, overlooking the capability to evaluate subjective questions which is extremely common for large language models. Additionally, these methods predominantly utilize centralized datasets for evaluation, with question banks concentrated within the evaluation platforms themselves. Moreover, the evaluation processes employed by these platforms often overlook personalized factors, neglecting to consider the individual characteristics of both the evaluators and the models being evaluated. To address these limitations, we propose a novel anonymous crowd-sourcing evaluation platform, BingJian, for large language models that employs a competitive scoring mechanism where users participate in ranking models based on their performance. This platform stands out not only for its support of centralized evaluations to assess the general capabilities of models but also for offering an open evaluation gateway. Through this gateway, users have the opportunity to submit their questions, testing the models on a personalized and potentially broader range of capabilities. Furthermore, our platform introduces personalized evaluation scenarios, leveraging various forms of human-computer interaction to assess large language models in a manner that accounts for individual user preferences and contexts. The demonstration of BingJian can be accessed at https://github.com/Mingyue-Cheng/Bingjian.

Zhuohao Yin, Xin Huang

Visual Word Sense Disambiguation (VWSD) is a multi-modal task that aims to select, among a batch of candidate images, the one that best entails the target word's meaning within a limited context. In this paper, we propose a multi-modal retrieval framework that maximally leverages pretrained Vision-Language models, as well as open knowledge bases and datasets. Our system consists of the following key components: (1) Gloss matching: a pretrained bi-encoder model is used to match contexts with proper senses of the target words; (2) Prompting: matched glosses and other textual information, such as synonyms, are incorporated using a prompting template; (3) Image retrieval: semantically matching images are retrieved from large open datasets using prompts as queries; (4) Modality fusion: contextual information from different modalities are fused and used for prediction. Although our system does not produce the most competitive results at SemEval-2023 Task 1, we are still able to beat nearly half of the teams. More importantly, our experiments reveal acute insights for the field of Word Sense Disambiguation (WSD) and multi-modal learning. Our code is available on GitHub.

Wenbo Shang, Xuliang Zhu, Xin Huang

Unified graph representation learning aims to generate node embeddings, which can be applied to multiple downstream applications of graph analytics. However, existing studies based on graph neural networks and language models either suffer from the limitations of numerous training needs toward specific downstream predictions, poor generalization, or shallow semantic features. In this work, we propose a novel Path-LLM model to efficiently learn unified graph representation, which leverages a powerful large language model (LLM) to incorporate our proposed path features. Our Path-LLM framework consists of four well-designed techniques. First, we develop a new mechanism of long-to-short shortest path (L2SP) selection, which can cover key connections between different dense groups. An in-depth analysis and comparison of different path selections is conducted to justify the rationale behind our designed L2SP method. Next, we design path textualization to obtain L2SP-based training texts with key phrase selection from node text attributes. We then feed the texts into a self-supervised LLM training process to align next node/edge generation in L2SP with next token generation in causal language modeling for graph representation learning and finally extract the unified graph embeddings. We theoretically analyze the algorithm complexity of our Path-LLM approach. Extensive experiments on large-scale graph benchmarks validate the superiority of Path-LLM against state-of-the-art methods WalkLM, GraphGPT, OFA, and GraphTranslator on two classical graph learning tasks (node classification and edge validation) and one NP-hard graph query processing task (keyword search). Compared with WalkLM, our approach saves more than 90% of training paths on millions-scale graphs and runs at most 35x faster.

Xin Huang, Jongryool Kim, Bradley Rees et al.

Graph neural networks (GNNs) have received great attention due to their success in various graph-related learning tasks. Several GNN frameworks have then been developed for fast and easy implementation of GNN models. Despite their popularity, they are not well documented, and their implementations and system performance have not been well understood. In particular, unlike the traditional GNNs that are trained based on the entire graph in a full-batch manner, recent GNNs have been developed with different graph sampling techniques for mini-batch training of GNNs on large graphs. While they improve the scalability, their training times still depend on the implementations in the frameworks as sampling and its associated operations can introduce non-negligible overhead and computational cost. In addition, it is unknown how much the frameworks are 'eco-friendly' from a green computing perspective. In this paper, we provide an in-depth study of two mainstream GNN frameworks along with three state-of-the-art GNNs to analyze their performance in terms of runtime and power/energy consumption. We conduct extensive benchmark experiments at several different levels and present detailed analysis results and observations, which could be helpful for further improvement and optimization.

Xixuan Yang, Xin Huang, Chiming Duan et al.

Anomaly detection is crucial for ensuring the stability and reliability of web service systems. Logs and metrics contain multiple information that can reflect the system's operational state and potential anomalies. Thus, existing anomaly detection methods use logs and metrics to detect web service systems' anomalies through data fusion approaches. They associate logs and metrics using coarse-grained time window alignment and capture the normal patterns of system operation through reconstruction. However, these methods have two issues that limit their performance in anomaly detection. First, due to asynchrony between logs and metrics, coarse-grained time window alignment cannot achieve a precise association between the two modalities. Second, reconstruction-based methods suffer from severe overgeneralization problems, resulting in anomalies being accurately reconstructed. In this paper, we propose a novel anomaly detection method named FFAD to address these two issues. On the one hand, FFAD employs graph-based alignment to mine and extract associations between the modalities from the constructed log-metric relation graph, achieving precise associations between logs and metrics. On the other hand, we improve the model's fit to normal data distributions through Fourier Frequency Focus, thereby enhancing the effectiveness of anomaly detection. We validated the effectiveness of our model on two real-world industrial datasets and one open-source dataset. The results show that our method achieves an average anomaly detection F1-score of 93.6%, representing an 8.8% improvement over previous state-of-the-art methods.

Runmin Jiang, Zhaoxin Fan, Junhao Wu et al.

3D medical image segmentation is a challenging task with crucial implications for disease diagnosis and treatment planning. Recent advances in deep learning have significantly enhanced fully supervised medical image segmentation. However, this approach heavily relies on labor-intensive and time-consuming fully annotated ground-truth labels, particularly for 3D volumes. To overcome this limitation, we propose a novel probabilistic-aware weakly supervised learning pipeline, specifically designed for 3D medical imaging. Our pipeline integrates three innovative components: a Probability-based Pseudo Label Generation technique for synthesizing dense segmentation masks from sparse annotations, a Probabilistic Multi-head Self-Attention network for robust feature extraction within our Probabilistic Transformer Network, and a Probability-informed Segmentation Loss Function to enhance training with annotation confidence. Demonstrating significant advances, our approach not only rivals the performance of fully supervised methods but also surpasses existing weakly supervised methods in CT and MRI datasets, achieving up to 18.1% improvement in Dice scores for certain organs. The code is available at https://github.com/runminjiang/PW4MedSeg.

Yuqi Chen, Weihan Zhang, Xin Huang

Public-private graph, where a public network is visible to everyone and every user is also associated with its own small private graph accessed by itself only, widely exists in real-world applications of social networks and financial networks. Most existing work on community search, finding a query-dependent community containing a given query, only studies on a public graph, neglecting the privacy issues in public-private networks. However, considering both the public and private attributes of users enables community search to be more accurate, comprehensive, and personalized to discover hidden patterns. In this paper, we study a novel problem of attributed community search in public-private graphs (ACS-PP), aiming to find a connected k-core community that shares the most keywords with the query node. This problem uncovers structurally cohesive communities, such as interest-based user groups or core teams in collaborative networks. To optimize search efficiency, we propose an integrated scheme of constructing a public global graph index and a private personalized graph index. For the private index, we developed a compact structure of the PP-FP-tree index. The PP-FP-tree is constructed based on the public and private neighbors of the query node in the public-private graph, serving as an efficient index to mine frequent node sets that share the most common attributes with the query node. Extensive experiments on real public-private graph datasets validate both the efficiency and quality of our proposed PP-FP search algorithm against existing competitors. The case study on public-private collaboration networks provides insights into the discovery of public-private communities.

Junxiao Liu, Zhijun Wang, Yixiao Li et al.

Long reasoning models often struggle in multilingual settings: they tend to reason in English for non-English questions; when constrained to reasoning in the question language, accuracies drop substantially. The struggle is caused by the limited abilities for both multilingual question understanding and multilingual reasoning. To address both problems, we propose TRIT (Translation-Reasoning Integrated Training), a self-improving framework that integrates the training of translation into multilingual reasoning. Without external feedback or additional multilingual data, our method jointly enhances multilingual question understanding and response generation. On MMATH, our method outperforms multiple baselines by an average of 7 percentage points, improving both answer correctness and language consistency. Further analysis reveals that integrating translation training improves cross-lingual question alignment by over 10 percentage points and enhances translation quality for both mathematical questions and general-domain text, with gains up to 8.4 COMET points on FLORES-200.

Xin Huang, Tarun Kumar Vangani, Minh Duc Pham et al.

Multilingual large language models (MLLMs) have shown impressive capabilities across a variety of languages. However, efficacy can differ greatly between different language families, especially for those with limited linguistic resources. This report presents MERaLiON-TextLLM, a series of open-source language models specifically tailored to improve understanding and generation in Chinese, Indonesian, Malay, and Singlish. The initial released model is built on Llama-3-8B-Base and refined through a meticulously crafted process of continued pre-training and weight merging. Our approach achieves performance improvements across benchmarks in these languages, exceeding the capabilities of the official Llama-3 models. We provide the model checkpoints as a resource to support further research and development in cross-lingual language understanding.

Weihua Zheng, Chang Liu, Zhengyuan Liu et al.

Multilingual Large Language Models (LLMs) struggle with cross-lingual tasks due to data imbalances between high-resource and low-resource languages, as well as monolingual bias in pre-training. Existing methods, such as bilingual fine-tuning and contrastive alignment, can improve cross-lingual performance, but they often require extensive parallel data or suffer from instability. To address these challenges, we introduce a Cross-Lingual Mapping Task during the pre-training phase, which enhances cross-lingual alignment without compromising monolingual fluency. Our approach bi-directionally maps languages within the LLM embedding space, improving both language generation and comprehension. We further propose a Language Alignment Coefficient to robustly quantify cross-lingual consistency, even in limited-data scenarios. Experimental results on machine translation (MT), cross-lingual natural language understanding (CLNLU), and cross-lingual question answering (CLQA) show that our model achieves gains of up to 11.9 BLEU points in MT, 6.72 points in CLQA BERTScore-Precision, and more than 5% in CLNLU accuracy over strong multilingual baselines. These findings highlight the potential of incorporating cross-lingual objectives into pre-training to improve multilingual LLMs.

Yunzhi Shen, Hao Zhou, Xin Huang et al.

Reinforcement learning (RL) has shown strong promise for LLM-based machine translation, with recent methods such as GRPO demonstrating notable gains; nevertheless, translation-oriented RL remains challenged by noisy learning signals arising from Monte Carlo return estimation, as well as a large trajectory space that favors global exploration over fine-grained local optimization. We introduce \textbf{PEGRL}, a \textit{two-stage} RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. At each iteration, translation outputs are sampled to construct post-editing inputs, allowing return estimation in the post-editing stage to benefit from conditioning on the current translation behavior, while jointly supporting both global exploration and fine-grained local optimization. A task-specific weighting scheme further balances the contributions of translation and post-editing objectives, yielding a biased yet more sample-efficient estimator. Experiments on English$\to$Finnish, English$\to$Turkish, and English$\leftrightarrow$Chinese show consistent gains over RL baselines, and for English$\to$Turkish, performance on COMET-KIWI is comparable to advanced LLM-based systems (DeepSeek-V3.2).

Min Wang, Xin Huang, Guoqing Zhou et al.

Neural Radiance Fields (NeRFs) have demonstrated prominent performance in novel view synthesis. However, their input heavily relies on image acquisition under normal light conditions, making it challenging to learn accurate scene representation in low-light environments where images typically exhibit significant noise and severe color distortion. To address these challenges, we propose a novel approach, Bright-NeRF, which learns enhanced and high-quality radiance fields from multi-view low-light raw images in an unsupervised manner. Our method simultaneously achieves color restoration, denoising, and enhanced novel view synthesis. Specifically, we leverage a physically-inspired model of the sensor's response to illumination and introduce a chromatic adaptation loss to constrain the learning of response, enabling consistent color perception of objects regardless of lighting conditions. We further utilize the raw data's properties to expose the scene's intensity automatically. Additionally, we have collected a multi-view low-light raw image dataset to advance research in this field. Experimental results demonstrate that our proposed method significantly outperforms existing 2D and 3D approaches. Our code and dataset will be made publicly available.

Charoes Huang, Xin Huang, Ngoc Phu Tran et al.

The Model Context Protocol (MCP) has rapidly emerged as a universal standard for connecting AI assistants to external tools and data sources. While MCP simplifies integration between AI applications and various services, it introduces significant security vulnerabilities, particularly on the client side. In this work we conduct threat modelings of MCP implementations using STRIDE (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege) and DREAD (Damage, Reproducibility, Exploitability, Affected Users, Discoverability) frameworks across five key components: (1) MCP Host and Client, (2) LLM, (3) MCP Server, (4) External Data Stores, and (5) Authorization Server. This comprehensive analysis reveals tool poisoning-where malicious instructions are embedded in tool metadata-as the most prevalent and impactful client-side vulnerability. We therefore focus our empirical evaluation on this critical attack vector, providing a systematic comparison of how seven major MCP clients validate and defend against tool poisoning attacks. Our analysis reveals significant security issues with most tested clients due to insufficient static validation and parameter visibility. We propose a multi-layered defense strategy encompassing static metadata analysis, model decision path tracking, behavioral anomaly detection, and user transparency mechanisms. This research addresses a critical gap in MCP security, which has primarily focused on server-side vulnerabilities, and provides actionable recommendations and mitigation strategies for securing AI agent ecosystems.

Charoes Huang, Xin Huang, Amin Milani Fard

The Model Context Protocol (MCP) has emerged as a standard for connecting Large Language Models (LLMs) to external tools and data. However, MCP servers often expose privileged capabilities, such as file system access, network requests, and command execution that can be exploited if not properly secured. We present mcp-sec-audit, an extensible security assessment toolkit designed specifically for MCP servers. It implements static pattern matching for Python-based MCP servers and dynamic sandboxed fuzzing and monitoring via Docker and eBPF. The tool detects risky capabilities through configurable rule-based analysis and provides mitigation recommendations.

Charoes Huang, Xin Huang, Amin Milani Fard

Prompt injection is listed as the number-one vulnerability class in the OWASP Top 10 for LLM Applications that can subvert LLM guardrails, disclose sensitive data, and trigger unauthorized tool use. Developers are rapidly adopting AI-assisted development tools built on the Model Context Protocol (MCP). However, their convenience comes with security risks, especially prompt-injection attacks delivered via tool-poisoning vectors. While prior research has studied prompt injection in LLMs, the security posture of real-world MCP clients remains underexplored. We present the first empirical analysis of prompt injection with the tool-poisoning vulnerability across seven widely used MCP clients: Claude Desktop, Claude Code, Cursor, Cline, Continue, Gemini CLI, and Langflow. We identify their detection and mitigation mechanisms, as well as the coverage of security features, including static validation, parameter visibility, injection detection, user warnings, execution sandboxing, and audit logging. Our evaluation reveals significant disparities. While some clients, such as Claude Desktop, implement strong guardrails, others, such as Cursor, exhibit high susceptibility to cross-tool poisoning, hidden parameter exploitation, and unauthorized tool invocation. We further provide actionable guidance for MCP implementers and the software engineering community seeking to build secure AI-assisted development workflows.

Xin Huang, Ruibin Li, Tong Jia et al.

Vision-Language Models (VLMs) are essential for multimodal tasks, especially compositional reasoning (CR) tasks, which require distinguishing fine-grained semantic differences between visual and textual embeddings. However, existing methods primarily fine-tune the model by generating text-based hard negative samples, neglecting the importance of image-based negative samples, which results in insufficient training of the visual encoder and ultimately impacts the overall performance of the model. Moreover, negative samples are typically treated uniformly, without considering their difficulty levels, and the alignment of positive samples is insufficient, which leads to challenges in aligning difficult sample pairs. To address these issues, we propose Adaptive Hard Negative Perturbation Learning (AHNPL). AHNPL translates text-based hard negatives into the visual domain to generate semantically disturbed image-based negatives for training the model, thereby enhancing its overall performance. AHNPL also introduces a contrastive learning approach using a multimodal hard negative loss to improve the model's discrimination of hard negatives within each modality and a dynamic margin loss that adjusts the contrastive margin according to sample difficulty to enhance the distinction of challenging sample pairs. Experiments on three public datasets demonstrate that our method effectively boosts VLMs' performance on complex CR tasks. The source code is available at https://github.com/nynu-BDAI/AHNPL.

Jihao Liu, Xin Huang, Jinliang Zheng et al.

This paper introduces MM-Instruct, a large-scale dataset of diverse and high-quality visual instruction data designed to enhance the instruction-following capabilities of large multimodal models (LMMs). While existing visual instruction datasets often focus on question-answering, they struggle to generalize to broader application scenarios such as creative writing, summarization, or image analysis. To address these limitations, we propose a novel approach to constructing MM-Instruct that leverages the strong instruction-following capabilities of existing LLMs to generate novel visual instruction data from large-scale but conventional image captioning datasets. MM-Instruct first leverages ChatGPT to automatically generate diverse instructions from a small set of seed instructions through augmenting and summarization. It then matches these instructions with images and uses an open-sourced large language model (LLM) to generate coherent answers to the instruction-image pairs. The LLM is grounded by the detailed text descriptions of images in the whole answer generation process to guarantee the alignment of the instruction data. Moreover, we introduce a benchmark based on the generated instruction data to evaluate the instruction-following capabilities of existing LMMs. We demonstrate the effectiveness of MM-Instruct by training a LLaVA-1.5 model on the generated data, denoted as LLaVA-Instruct, which exhibits significant improvements in instruction-following capabilities compared to LLaVA-1.5 models. The MM-Instruct dataset, benchmark, and pre-trained models are available at https://github.com/jihaonew/MM-Instruct.

Peng Hu, Sizhe Liu, Changjiang Gao et al.

Large Language Models have demonstrated impressive reasoning capabilities across multiple languages. However, the relationship between capabilities in different languages is less explored. In this work, we decompose the process of reasoning tasks into two separated components: knowledge retrieval and knowledge-free reasoning, and analyze the relationship between cross-lingual transferability and these two components. With adapted commonsense reasoning datasets and constructed knowledge-free reasoning datasets, we show that the knowledge-free reasoning capability can be nearly perfectly transferred across various source-target language directions despite the secondary impact of resource in some specific target languages, while cross-lingual knowledge retrieval significantly hinders the transfer. Moreover, by analyzing the hidden states and feed-forward network neuron activation during the reasoning, we show that higher similarity of hidden representations and larger overlap of activated neurons could explain the better cross-lingual transferability of knowledge-free reasoning than knowledge retrieval. Thus, we hypothesize that knowledge-free reasoning shares similar neurons in different languages for reasoning, while knowledge is stored separately in different languages. Our code and data is available at: https://github.com/NJUNLP/Knowledge-Free-Reasoning.

Haolong Yan, Kaijun Tan, Yeqing Shen et al.

We investigate a critical yet under-explored question in Large Vision-Language Models (LVLMs): Do LVLMs genuinely comprehend interleaved image-text in the document? Existing document understanding benchmarks often assess LVLMs using question-answer formats, which are information-sparse and difficult to guarantee the coverage of long-range dependencies. To address this issue, we introduce a novel and challenging Multimodal Document Summarization Benchmark (M-DocSum-Bench), which comprises 500 high-quality arXiv papers, along with interleaved multimodal summaries aligned with human preferences. M-DocSum-Bench is a reference-based generation task and necessitates the generation of interleaved image-text summaries using provided reference images, thereby simultaneously evaluating capabilities in understanding, reasoning, localization, and summarization within complex multimodal document scenarios. To facilitate this benchmark, we develop an automated framework to construct summaries and propose a fine-grained evaluation method called M-DocEval. Moreover, we further develop a robust summarization baseline, i.e., M-DocSum-7B, by progressive two-stage training with diverse instruction and preference data. The extensive results on our M-DocSum-Bench reveal that the leading LVLMs struggle to maintain coherence and accurately integrate information within long and interleaved contexts, often exhibiting confusion between similar images and a lack of robustness. Notably, M-DocSum-7B achieves state-of-the-art performance compared to larger and closed-source models (including GPT-4o, Gemini Pro, Claude-3.5-Sonnet and Qwen2.5-VL-72B, etc.), demonstrating the potential of LVLMs for improved interleaved image-text understanding. The code, data, and models are available at https://github.com/stepfun-ai/M-DocSum-Bench.

Xuejiao Tang, Xin Huang, Wenbin Zhang et al.

Visual Commonsense Reasoning (VCR) predicts an answer with corresponding rationale, given a question-image input. VCR is a recently introduced visual scene understanding task with a wide range of applications, including visual question answering, automated vehicle systems, and clinical decision support. Previous approaches to solving the VCR task generally rely on pre-training or exploiting memory with long dependency relationship encoded models. However, these approaches suffer from a lack of generalizability and prior knowledge. In this paper we propose a dynamic working memory based cognitive VCR network, which stores accumulated commonsense between sentences to provide prior knowledge for inference. Extensive experiments show that the proposed model yields significant improvements over existing methods on the benchmark VCR dataset. Moreover, the proposed model provides intuitive interpretation into visual commonsense reasoning. A Python implementation of our mechanism is publicly available at https://github.com/tanjatang/DMVCR

Xin Huang, Xinxin Wang, Wenyu Lv et al.

Being effective and efficient is essential to an object detector for practical use. To meet these two concerns, we comprehensively evaluate a collection of existing refinements to improve the performance of PP-YOLO while almost keep the infer time unchanged. This paper will analyze a collection of refinements and empirically evaluate their impact on the final model performance through incremental ablation study. Things we tried that didn't work will also be discussed. By combining multiple effective refinements, we boost PP-YOLO's performance from 45.9% mAP to 49.5% mAP on COCO2017 test-dev. Since a significant margin of performance has been made, we present PP-YOLOv2. In terms of speed, PP-YOLOv2 runs in 68.9FPS at 640x640 input size. Paddle inference engine with TensorRT, FP16-precision, and batch size = 1 further improves PP-YOLOv2's infer speed, which achieves 106.5 FPS. Such a performance surpasses existing object detectors with roughly the same amount of parameters (i.e., YOLOv4-CSP, YOLOv5l). Besides, PP-YOLOv2 with ResNet101 achieves 50.3% mAP on COCO2017 test-dev. Source code is at https://github.com/PaddlePaddle/PaddleDetection.

Jianjun Wei, Yue Liu, Xin Huang et al.

This paper explores the applications and challenges of graph neural networks (GNNs) in processing complex graph data brought about by the rapid development of the Internet. Given the heterogeneity and redundancy problems that graph data often have, traditional GNN methods may be overly dependent on the initial structure and attribute information of the graph, which limits their ability to accurately simulate more complex relationships and patterns in the graph. Therefore, this study proposes a graph neural network model under a self-supervised learning framework, which can flexibly combine different types of additional information of the attribute graph and its nodes, so as to better mine the deep features in the graph data. By introducing a self-supervisory mechanism, it is expected to improve the adaptability of existing models to the diversity and complexity of graph data and improve the overall performance of the model.