Jie Li

Sicong Wang, Ruiting Dong, Yue Liu et al.

Real-world user behavior rarely consists of isolated actions; instead, it often forms intent flows governed by spatiotemporal dependencies. To provide integrated service recommendations, we focus on the task of Generative Spatiotemporal Intent Sequence Recommendation (GSISR), which aims to generate intent sequences that are logically coherent and physically executable within complex spatiotemporal contexts. While LLMs offer strong reasoning potential for GSISR, direct industrial deployment is limited by high inference latency and context-mismatched or physically infeasible plans. To address these challenges, we propose a generative framework, GPlan, that internalizes LLM reasoning into lightweight models through two components. First, to enable reasoning under strict latency constraints, we introduce Progressive Implicit CoT Distillation, which compresses explicit reasoning processes into reserved latent tokens, allowing small models to inherit complex planning logic without generating long reasoning text. Second, to address the disconnect between general knowledge and real-world constraints, we design Spatiotemporal Counterfactual DPO. By aligning the model with counterfactual context-plan pairs, we improve sensitivity to spatiotemporal context and reduce context-mismatched plans. Offline experiments and online A/B testing demonstrate that our approach improves sequence coherence and context responsiveness. Our implementation and the anonymized GSISR dataset are available at https://github.com/alibaba/GPlan.

Katie Z Luo, Xinshuo Weng, Yan Wang et al.

Autonomous driving has traditionally relied heavily on costly and labor-intensive High Definition (HD) maps, hindering scalability. In contrast, Standard Definition (SD) maps are more affordable and have worldwide coverage, offering a scalable alternative. In this work, we systematically explore the effect of SD maps for real-time lane-topology understanding. We propose a novel framework to integrate SD maps into online map prediction and propose a Transformer-based encoder, SD Map Encoder Representations from transFormers, to leverage priors in SD maps for the lane-topology prediction task. This enhancement consistently and significantly boosts (by up to 60%) lane detection and topology prediction on current state-of-the-art online map prediction methods without bells and whistles and can be immediately incorporated into any Transformer-based lane-topology method. Code is available at https://github.com/NVlabs/SMERF.

Jie Li, Wenzhao Wu, Junqi Hu et al.

While large language models (LLMs) have been extensively evaluated on code generation tasks for general-purpose programming and GPU-accelerated environments (e.g., PyTorch, CUDA), their capabilities in CPU-oriented high-performance computing (HPC) across diverse architectures remain underexplored. To bridge this gap, we introduce CodegenBench, a comprehensive benchmark suite designed to evaluate the generation of efficient parallel code across three distinct hardware platforms: x86_64, Sunway, and Kunpeng. Our benchmark comprises 106 standard Basic Linear Algebra Subprograms (BLAS) routines establishing a fundamental baseline, alongside 20 specialized computational kernels adapted for each of the unique supercomputing architectures (LeetSunway and LeetKunpeng). Our extensive evaluation reveals that while state-of-the-art LLMs can generate optimized code for ubiquitous architectures like x86_64, they exhibit significant performance degradation on domain-specific architectures with limited public documentation and training data, highlighting critical limitations in cross-platform generalization. Furthermore, our analysis of factors influencing code quality such as implementation length and task complexity indicates that current LLMs are most effective for moderately difficult problems requiring concise code snippets. We open-source our dataset and automated evaluation infrastructure to facilitate future research in LLM-driven high-performance code generation. The resources are available at https://anonymous.4open.science/r/CodegenBench-EDE1/ and https://anonymous.4open.science/r/CodegenBenchDataset-2551.

Ziqi Pang, Jie Li, Pavel Tokmakov et al.

This work proposes an end-to-end multi-camera 3D multi-object tracking (MOT) framework. It emphasizes spatio-temporal continuity and integrates both past and future reasoning for tracked objects. Thus, we name it "Past-and-Future reasoning for Tracking" (PF-Track). Specifically, our method adapts the "tracking by attention" framework and represents tracked instances coherently over time with object queries. To explicitly use historical cues, our "Past Reasoning" module learns to refine the tracks and enhance the object features by cross-attending to queries from previous frames and other objects. The "Future Reasoning" module digests historical information and predicts robust future trajectories. In the case of long-term occlusions, our method maintains the object positions and enables re-association by integrating motion predictions. On the nuScenes dataset, our method improves AMOTA by a large margin and remarkably reduces ID-Switches by 90% compared to prior approaches, which is an order of magnitude less. The code and models are made available at https://github.com/TRI-ML/PF-Track.

Dian Chen, Jie Li, Vitor Guizilini et al.

3D object detection from visual sensors is a cornerstone capability of robotic systems. State-of-the-art methods focus on reasoning and decoding object bounding boxes from multi-view camera input. In this work we gain intuition from the integral role of multi-view consistency in 3D scene understanding and geometric learning. To this end, we introduce VEDet, a novel 3D object detection framework that exploits 3D multi-view geometry to improve localization through viewpoint awareness and equivariance. VEDet leverages a query-based transformer architecture and encodes the 3D scene by augmenting image features with positional encodings from their 3D perspective geometry. We design view-conditioned queries at the output level, which enables the generation of multiple virtual frames during training to learn viewpoint equivariance by enforcing multi-view consistency. The multi-view geometry injected at the input level as positional encodings and regularized at the loss level provides rich geometric cues for 3D object detection, leading to state-of-the-art performance on the nuScenes benchmark. The code and model are made available at https://github.com/TRI-ML/VEDet.

Xinxu Ge, Xin Liu, Zitong Yu et al.

Face anti-spoofing (FAS) plays a vital role in preventing face recognition (FR) systems from presentation attacks. Nowadays, FAS systems face the challenge of domain shift, impacting the generalization performance of existing FAS methods. In this paper, we rethink about the inherence of domain shift and deconstruct it into two factors: image style and image quality. Quality influences the purity of the presentation of spoof information, while style affects the manner in which spoof information is presented. Based on our analysis, we propose DiffFAS framework, which quantifies quality as prior information input into the network to counter image quality shift, and performs diffusion-based high-fidelity cross-domain and cross-attack types generation to counter image style shift. DiffFAS transforms easily collectible live faces into high-fidelity attack faces with precise labels while maintaining consistency between live and spoof face identities, which can also alleviate the scarcity of labeled data with novel type attacks faced by nowadays FAS system. We demonstrate the effectiveness of our framework on challenging cross-domain and cross-attack FAS datasets, achieving the state-of-the-art performance. Available at https://github.com/murphytju/DiffFAS.

Jiaqi Yu, Xin Wang, Yixu Wang et al.

Large language models increasingly stream long, reasoning-intensive responses in real time, making when to moderate as critical as whether to moderate. Existing guardrails fall into two unsatisfactory extremes: response-level methods delay intervention until the full output is generated, whereas token-level methods act on incomplete semantics, often producing unstable decisions and excessive guard invocations. To address this challenge, we propose SentGuard, a sentence-level streaming guardrail that operates in parallel with generation. A lightweight waiting buffer groups streamed tokens into sentence chunks and releases only verified chunks to the user, introducing a small offset that enables SentGuard to assess the current prefix while the target LLM decodes subsequent content. To support this, we construct StreamSafe, a benchmark with structured per-sentence annotations across 8 harm categories, capturing the evolution of safety risks across both reasoning and response segments. We further train SentGuard with a coarse-to-fine objective to detect unsafe intent as soon as it emerges at sentence boundaries. Experiments on 5 safety benchmarks show that SentGuard outperforms existing baselines, detecting 90.5% of unsafe cases within two sentences while maintaining a low streaming false-positive rate of 7.41%.

Jingyuan Yang, Jie Li, Leida Li et al.

Visual Emotion Analysis (VEA), which aims to predict people's emotions towards different visual stimuli, has become an attractive research topic recently. Rather than a single label classification task, it is more rational to regard VEA as a Label Distribution Learning (LDL) problem by voting from different individuals. Existing methods often predict visual emotion distribution in a unified network, neglecting the inherent subjectivity in its crowd voting process. In psychology, the \textit{Object-Appraisal-Emotion} model has demonstrated that each individual's emotion is affected by his/her subjective appraisal, which is further formed by the affective memory. Inspired by this, we propose a novel \textit{Subjectivity Appraise-and-Match Network (SAMNet)} to investigate the subjectivity in visual emotion distribution. To depict the diversity in crowd voting process, we first propose the \textit{Subjectivity Appraising} with multiple branches, where each branch simulates the emotion evocation process of a specific individual. Specifically, we construct the affective memory with an attention-based mechanism to preserve each individual's unique emotional experience. A subjectivity loss is further proposed to guarantee the divergence between different individuals. Moreover, we propose the \textit{Subjectivity Matching} with a matching loss, aiming at assigning unordered emotion labels to ordered individual predictions in a one-to-one correspondence with the Hungarian algorithm. Extensive experiments and comparisons are conducted on public visual emotion distribution datasets, and the results demonstrate that the proposed SAMNet consistently outperforms the state-of-the-art methods. Ablation study verifies the effectiveness of our method and visualization proves its interpretability.

Lei Han, Qingxu Zhu, Jiapeng Sheng et al.

Knowledge from animals and humans inspires robotic innovations. Numerous efforts have been made to achieve agile locomotion in quadrupedal robots through classical controllers or reinforcement learning approaches. These methods usually rely on physical models or handcrafted rewards to accurately describe the specific system, rather than on a generalized understanding like animals do. Here we propose a hierarchical framework to construct primitive-, environmental- and strategic-level knowledge that are all pre-trainable, reusable and enrichable for legged robots. The primitive module summarizes knowledge from animal motion data, where, inspired by large pre-trained models in language and image understanding, we introduce deep generative models to produce motor control signals stimulating legged robots to act like real animals. Then, we shape various traversing capabilities at a higher level to align with the environment by reusing the primitive module. Finally, a strategic module is trained focusing on complex downstream tasks by reusing the knowledge from previous levels. We apply the trained hierarchical controllers to the MAX robot, a quadrupedal robot developed in-house, to mimic animals, traverse complex obstacles and play in a designed challenging multi-agent chase tag game, where lifelike agility and strategy emerge in the robots.

Adam W. Harley, Zhaoyuan Fang, Jie Li et al.

Building 3D perception systems for autonomous vehicles that do not rely on high-density LiDAR is a critical research problem because of the expense of LiDAR systems compared to cameras and other sensors. Recent research has developed a variety of camera-only methods, where features are differentiably "lifted" from the multi-camera images onto the 2D ground plane, yielding a "bird's eye view" (BEV) feature representation of the 3D space around the vehicle. This line of work has produced a variety of novel "lifting" methods, but we observe that other details in the training setups have shifted at the same time, making it unclear what really matters in top-performing methods. We also observe that using cameras alone is not a real-world constraint, considering that additional sensors like radar have been integrated into real vehicles for years already. In this paper, we first of all attempt to elucidate the high-impact factors in the design and training protocol of BEV perception models. We find that batch size and input resolution greatly affect performance, while lifting strategies have a more modest effect -- even a simple parameter-free lifter works well. Second, we demonstrate that radar data can provide a substantial boost to performance, helping to close the gap between camera-only and LiDAR-enabled systems. We analyze the radar usage details that lead to good performance, and invite the community to re-consider this commonly-neglected part of the sensor platform.

Ruiyang Xia, Decheng Liu, Jie Li et al.

Advanced manipulation techniques have provided criminals with opportunities to make social panic or gain illicit profits through the generation of deceptive media, such as forged face images. In response, various deepfake detection methods have been proposed to assess image authenticity. Sequential deepfake detection, which is an extension of deepfake detection, aims to identify forged facial regions with the correct sequence for recovery. Nonetheless, due to the different combinations of spatial and sequential manipulations, forged face images exhibit substantial discrepancies that severely impact detection performance. Additionally, the recovery of forged images requires knowledge of the manipulation model to implement inverse transformations, which is difficult to ascertain as relevant techniques are often concealed by attackers. To address these issues, we propose Multi-Collaboration and Multi-Supervision Network (MMNet) that handles various spatial scales and sequential permutations in forged face images and achieve recovery without requiring knowledge of the corresponding manipulation method. Furthermore, existing evaluation metrics only consider detection accuracy at a single inferring step, without accounting for the matching degree with ground-truth under continuous multiple steps. To overcome this limitation, we propose a novel evaluation metric called Complete Sequence Matching (CSM), which considers the detection accuracy at multiple inferring steps, reflecting the ability to detect integrally forged sequences. Extensive experiments on several typical datasets demonstrate that MMNet achieves state-of-the-art detection performance and independent recovery performance.

Tara Sadjadpour, Jie Li, Rares Ambrus et al.

Multi-object tracking is a cornerstone capability of any robotic system. The quality of tracking is largely dependent on the quality of the detector used. In many applications, such as autonomous vehicles, it is preferable to over-detect objects to avoid catastrophic outcomes due to missed detections. As a result, current state-of-the-art 3D detectors produce high rates of false-positives to ensure a low number of false-negatives. This can negatively affect tracking by making data association and track lifecycle management more challenging. Additionally, occasional false-negative detections due to difficult scenarios like occlusions can harm tracking performance. To address these issues in a unified framework, we propose to learn shape and spatio-temporal affinities between tracks and detections in consecutive frames. Our affinity provides a probabilistic matching that leads to robust data association, track lifecycle management, false-positive elimination, false-negative propagation, and sequential track confidence refinement. Though past 3D MOT approaches address a subset of components in this problem domain, we offer the first self-contained framework that addresses all these aspects of the 3D MOT problem. We quantitatively evaluate our method on the nuScenes tracking benchmark where we achieve 1st place amongst LiDAR-only trackers using CenterPoint detections. Our method estimates accurate and precise tracks, while decreasing the overall number of false-positive and false-negative tracks and increasing the number of true-positive tracks. We analyze our performance with 5 metrics, giving a comprehensive overview of our approach to indicate how our tracking framework may impact the ultimate goal of an autonomous mobile agent. We also present ablative experiments and qualitative results that demonstrate our framework's capabilities in complex scenarios.

Colton Stearns, Davis Rempe, Jie Li et al.

3D multi-object tracking aims to uniquely and consistently identify all mobile entities through time. Despite the rich spatiotemporal information available in this setting, current 3D tracking methods primarily rely on abstracted information and limited history, e.g. single-frame object bounding boxes. In this work, we develop a holistic representation of traffic scenes that leverages both spatial and temporal information of the actors in the scene. Specifically, we reformulate tracking as a spatiotemporal problem by representing tracked objects as sequences of time-stamped points and bounding boxes over a long temporal history. At each timestamp, we improve the location and motion estimates of our tracked objects through learned refinement over the full sequence of object history. By considering time and space jointly, our representation naturally encodes fundamental physical priors such as object permanence and consistency across time. Our spatiotemporal tracking framework achieves state-of-the-art performance on the Waymo and nuScenes benchmarks.

Jie Li, Paul Fearnhead, Piotr Fryzlewicz et al.

Detecting change-points in data is challenging because of the range of possible types of change and types of behaviour of data when there is no change. Statistically efficient methods for detecting a change will depend on both of these features, and it can be difficult for a practitioner to develop an appropriate detection method for their application of interest. We show how to automatically generate new offline detection methods based on training a neural network. Our approach is motivated by many existing tests for the presence of a change-point being representable by a simple neural network, and thus a neural network trained with sufficient data should have performance at least as good as these methods. We present theory that quantifies the error rate for such an approach, and how it depends on the amount of training data. Empirical results show that, even with limited training data, its performance is competitive with the standard CUSUM-based classifier for detecting a change in mean when the noise is independent and Gaussian, and can substantially outperform it in the presence of auto-correlated or heavy-tailed noise. Our method also shows strong results in detecting and localising changes in activity based on accelerometer data.

Hao Lin, Ke Wu, Jie Li et al.

Distributed learning is commonly used for training deep learning models, especially large models. In distributed learning, manual parallelism (MP) methods demand considerable human effort and have limited flexibility. Hence, automatic parallelism (AP) methods have recently been proposed for automating the parallel strategy optimization process. Existing AP methods suffer from sub-optimal solutions because they do not jointly optimize the two categories of parallel strategies (i.e., inter-layer parallelism and intra-layer parallelism). In this paper, we propose a novel AP method called UniAP, which unifies inter- and intra-layer automatic parallelism by mixed integer quadratic programming. To the best of our knowledge, UniAP is the first parallel method that can jointly optimize the two categories of parallel strategies to find an optimal solution. Experimental results show that UniAP outperforms state-of-the-art methods by up to 3.80$\times$ in throughput and reduces strategy optimization time by up to 107$\times$ across five Transformer-based models.

Yukang Zhang, Yan Yan, Jie Li et al.

Visible-infrared person re-identification (VI-ReID), which aims to search identities across different spectra, is a challenging task due to large cross-modality discrepancy between visible and infrared images. The key to reduce the discrepancy is to filter out identity-irrelevant interference and effectively learn modality-invariant person representations. In this paper, we propose a novel Modality Restitution and Compensation Network (MRCN) to narrow the gap between the two modalities. Specifically, we first reduce the modality discrepancy by using two Instance Normalization (IN) layers. Next, to reduce the influence of IN layers on removing discriminative information and to reduce modality differences, we propose a Modality Restitution Module (MRM) and a Modality Compensation Module (MCM) to respectively distill modality-irrelevant and modality-relevant features from the removed information. Then, the modality-irrelevant features are used to restitute to the normalized visible and infrared features, while the modality-relevant features are used to compensate for the features of the other modality. Furthermore, to better disentangle the modality-relevant features and the modality-irrelevant features, we propose a novel Center-Quadruplet Causal (CQC) loss to encourage the network to effectively learn the modality-relevant features and the modality-irrelevant features. Extensive experiments are conducted to validate the superiority of our method on the challenging SYSU-MM01 and RegDB datasets. More remarkably, our method achieves 95.1% in terms of Rank-1 and 89.2% in terms of mAP on the RegDB dataset.

Shuman Fang, Shuai Liu, Jie Li et al.

Human-Object Interaction (HOI) detection aims to understand the interactions between humans and objects, which plays a curtail role in high-level semantic understanding tasks. However, most works pursue designing better architectures to learn overall features more efficiently, while ignoring the long-tail nature of interaction-object pair categories. In this paper, we propose to alleviate the impact of such an unbalanced distribution via Virtual Image Leaning (VIL). Firstly, a novel label-to-image approach, Multiple Steps Image Creation (MUSIC), is proposed to create a high-quality dataset that has a consistent distribution with real images. In this stage, virtual images are generated based on prompts with specific characterizations and selected by multi-filtering processes. Secondly, we use both virtual and real images to train the model with the teacher-student framework. Considering the initial labels of some virtual images are inaccurate and inadequate, we devise an Adaptive Matching-and-Filtering (AMF) module to construct pseudo-labels. Our method is independent of the internal structure of HOI detectors, so it can be combined with off-the-shelf methods by training merely 10 additional epochs. With the assistance of our method, multiple methods obtain significant improvements, and new state-of-the-art results are achieved on two benchmarks.

Li Li, Qingqing Li, Guozheng Xu et al.

Roof plane segmentation from airborne LiDAR point clouds is an important technology for 3D building model reconstruction. One of the key issues of plane segmentation is how to design powerful features that can exactly distinguish adjacent planar patches. The quality of point feature directly determines the accuracy of roof plane segmentation. Most of existing approaches use handcrafted features to extract roof planes. However, the abilities of these features are relatively low, especially in boundary area. To solve this problem, we propose a boundary-aware point clustering approach in Euclidean and embedding spaces constructed by a multi-task deep network for roof plane segmentation. We design a three-branch network to predict semantic labels, point offsets and extract deep embedding features. In the first branch, we classify the input data as non-roof, boundary and plane points. In the second branch, we predict point offsets for shifting each point toward its respective instance center. In the third branch, we constrain that points of the same plane instance should have the similar embeddings. We aim to ensure that points of the same plane instance are close as much as possible in both Euclidean and embedding spaces. However, although deep network has strong feature representative ability, it is still hard to accurately distinguish points near plane instance boundary. Therefore, we first group plane points into many clusters in the two spaces, and then we assign the rest boundary points to their closest clusters to generate final complete roof planes. In this way, we can effectively reduce the influence of unreliable boundary points. In addition, we prepare a synthetic dataset and two real datasets to train and evaluate our approach. The experiments results show that the proposed approach significantly outperforms the existing state-of-the-art approaches.

Yuxin Zhu, Xilei Zhu, Huiyu Duan et al.

Visual saliency prediction for omnidirectional videos (ODVs) has shown great significance and necessity for omnidirectional videos to help ODV coding, ODV transmission, ODV rendering, etc.. However, most studies only consider visual information for ODV saliency prediction while audio is rarely considered despite its significant influence on the viewing behavior of ODV. This is mainly due to the lack of large-scale audio-visual ODV datasets and corresponding analysis. Thus, in this paper, we first establish the largest audio-visual saliency dataset for omnidirectional videos (AVS-ODV), which comprises the omnidirectional videos, audios, and corresponding captured eye-tracking data for three video sound modalities including mute, mono, and ambisonics. Then we analyze the visual attention behavior of the observers under various omnidirectional audio modalities and visual scenes based on the AVS-ODV dataset. Furthermore, we compare the performance of several state-of-the-art saliency prediction models on the AVS-ODV dataset and construct a new benchmark. Our AVS-ODV datasets and the benchmark will be released to facilitate future research.

Guozhang Li, De Cheng, Xinpeng Ding et al.

Weakly Supervised Temporal Action Localization (WTAL) aims to classify and localize temporal boundaries of actions for the video, given only video-level category labels in the training datasets. Due to the lack of boundary information during training, existing approaches formulate WTAL as a classificationproblem, i.e., generating the temporal class activation map (T-CAM) for localization. However, with only classification loss, the model would be sub-optimized, i.e., the action-related scenes are enough to distinguish different class labels. Regarding other actions in the action-related scene ( i.e., the scene same as positive actions) as co-scene actions, this sub-optimized model would misclassify the co-scene actions as positive actions. To address this misclassification, we propose a simple yet efficient method, named bidirectional semantic consistency constraint (Bi-SCC), to discriminate the positive actions from co-scene actions. The proposed Bi-SCC firstly adopts a temporal context augmentation to generate an augmented video that breaks the correlation between positive actions and their co-scene actions in the inter-video; Then, a semantic consistency constraint (SCC) is used to enforce the predictions of the original video and augmented video to be consistent, hence suppressing the co-scene actions. However, we find that this augmented video would destroy the original temporal context. Simply applying the consistency constraint would affect the completeness of localized positive actions. Hence, we boost the SCC in a bidirectional way to suppress co-scene actions while ensuring the integrity of positive actions, by cross-supervising the original and augmented videos. Finally, our proposed Bi-SCC can be applied to current WTAL approaches, and improve their performance. Experimental results show that our approach outperforms the state-of-the-art methods on THUMOS14 and ActivityNet.

Jingyu Song, Xudong Chen, Liupei Lu et al.

High-definition (HD) maps provide environmental information for autonomous driving systems and are essential for safe planning. While existing methods with single-frame input achieve impressive performance for online vectorized HD map construction, they still struggle with complex scenarios and occlusions. We propose MemFusionMap, a novel temporal fusion model with enhanced temporal reasoning capabilities for online HD map construction. Specifically, we contribute a working memory fusion module that improves the model's memory capacity to reason across a history of frames. We also design a novel temporal overlap heatmap to explicitly inform the model about the temporal overlap information and vehicle trajectory in the Bird's Eye View space. By integrating these two designs, MemFusionMap significantly outperforms existing methods while also maintaining a versatile design for scalability. We conduct extensive evaluation on open-source benchmarks and demonstrate a maximum improvement of 5.4% in mAP over state-of-the-art methods. The project page for MemFusionMap is https://song-jingyu.github.io/MemFusionMap

Zongmin Liu, Jirui Wang, Jie Li et al.

With the expansive aging of global population, service robot with living assistance applied in indoor scenes will serve as a crucial role in the field of elderly care and health in the future. Service robots need to detect multiple targets when completing auxiliary tasks. However, indoor scenes are usually complex and there are many types of interference factors, leading to great challenges in the multiple targets detection. To overcome this technical difficulty, a novel improved Mask RCNN method for multiple targets detection in the indoor complex scenes is proposed in this paper. The improved model utilizes Mask RCNN as the network framework. On this basis, Convolutional Block Attention Module (CBAM) with channel mechanism and space mechanism is integrated, and the influence of different background, distance, angle and interference factors are comprehensively considered. Meanwhile, in order to evaluate the detection and identification effects of the established model, a comprehensive evaluation system based on loss function and Mean Average Precision (mAP) is established. For verification, experiments on the detection and identification effects under different distances, backgrounds, angles and interference factors were conducted. The results show that designed model improves the accuracy to a higher level and has a better anti-interference ability than other methods when the detection speed was nearly the same.

Laiyan Ding, Hualie Jiang, Jie Li et al.

Depth estimation is a cornerstone for autonomous driving, yet acquiring per-pixel depth ground truth for supervised learning is challenging. Self-Supervised Surround Depth Estimation (SSSDE) from consecutive images offers an economical alternative. While previous SSSDE methods have proposed different mechanisms to fuse information across images, few of them explicitly consider the cross-view constraints, leading to inferior performance, particularly in overlapping regions. This paper proposes an efficient and consistent pose estimation design and two loss functions to enhance cross-view consistency for SSSDE. For pose estimation, we propose to use only front-view images to reduce training memory and sustain pose estimation consistency. The first loss function is the dense depth consistency loss, which penalizes the difference between predicted depths in overlapping regions. The second one is the multi-view reconstruction consistency loss, which aims to maintain consistency between reconstruction from spatial and spatial-temporal contexts. Additionally, we introduce a novel flipping augmentation to improve the performance further. Our techniques enable a simple neural model to achieve state-of-the-art performance on the DDAD and nuScenes datasets. Last but not least, our proposed techniques can be easily applied to other methods. The code is available at https://github.com/denyingmxd/CVCDepth.

Kai Han, Xiaohu Huang, Yandong Li et al.

Existing machine learning models demonstrate excellent performance in image object recognition after training on a large-scale dataset under full supervision. However, these models only learn to map an image to a predefined class index, without revealing the actual semantic meaning of the object in the image. In contrast, vision-language models like CLIP are able to assign semantic class names to unseen objects in a 'zero-shot' manner, though they are once again provided a pre-defined set of candidate names at test-time. In this paper, we reconsider the recognition problem and task a vision-language model with assigning class names to images given only a large (essentially unconstrained) vocabulary of categories as prior information. We leverage non-parametric methods to establish meaningful relationships between images, allowing the model to automatically narrow down the pool of candidate names. Our proposed approach entails iteratively clustering the data and employing a voting mechanism to determine the most suitable class names. Additionally, we investigate the potential of incorporating additional textual features to enhance clustering performance. To achieve this, we employ the CLIP vision and text encoders to retrieve relevant texts from an external database, which can provide supplementary semantic information to inform the clustering process. Furthermore, we tackle this problem both in unsupervised and partially supervised settings, as well as with a coarse-grained and fine-grained search space as the unconstrained dictionary. Remarkably, our method leads to a roughly 50% improvement over the baseline on ImageNet in the unsupervised setting.

Pavel Tokmakov, Jie Li, Adrien Gaidon

The appearance of an object can be fleeting when it transforms. As eggs are broken or paper is torn, their color, shape and texture can change dramatically, preserving virtually nothing of the original except for the identity itself. Yet, this important phenomenon is largely absent from existing video object segmentation (VOS) benchmarks. In this work, we close the gap by collecting a new dataset for Video Object Segmentation under Transformations (VOST). It consists of more than 700 high-resolution videos, captured in diverse environments, which are 21 seconds long on average and densely labeled with instance masks. A careful, multi-step approach is adopted to ensure that these videos focus on complex object transformations, capturing their full temporal extent. We then extensively evaluate state-of-the-art VOS methods and make a number of important discoveries. In particular, we show that existing methods struggle when applied to this novel task and that their main limitation lies in over-reliance on static appearance cues. This motivates us to propose a few modifications for the top-performing baseline that improve its capabilities by better modeling spatio-temporal information. But more broadly, the hope is to stimulate discussion on learning more robust video object representations.

Dennis Park, Jie Li, Dian Chen et al.

A key contributor to recent progress in 3D detection from single images is monocular depth estimation. Existing methods focus on how to leverage depth explicitly, by generating pseudo-pointclouds or providing attention cues for image features. More recent works leverage depth prediction as a pretraining task and fine-tune the depth representation while training it for 3D detection. However, the adaptation is insufficient and is limited in scale by manual labels. In this work, we propose to further align depth representation with the target domain in unsupervised fashions. Our methods leverage commonly available LiDAR or RGB videos during training time to fine-tune the depth representation, which leads to improved 3D detectors. Especially when using RGB videos, we show that our two-stage training by first generating pseudo-depth labels is critical because of the inconsistency in loss distribution between the two tasks. With either type of reference data, our multi-task learning approach improves over the state of the art on both KITTI and NuScenes, while matching the test-time complexity of its single task sub-network.

Angen Ye, Boyuan Wang, Chaojun Ni et al.

World-Action Models (WAM) initialized from pre-trained video generation backbones have demonstrated remarkable potential for robot policy learning. However, existing approaches face two critical bottlenecks that hinder performance and deployment. First, jointly reasoning over future visual dynamics and corresponding actions incurs substantial inference overhead. Second, joint modeling often entangles visual and motion representations, making motion prediction accuracy heavily dependent on the quality of future video forecasts. To address these issues, we introduce GigaWorld-Policy, an action-centered WAM that learns 2D pixel-action dynamics while enabling efficient action decoding, with optional video generation. Specifically, we formulate policy training into two coupled components: the model predicts future action sequences conditioned on the current observation, and simultaneously generates future videos conditioned on the predicted actions and the same observation. The policy is supervised by both action prediction and video generation, providing richer learning signals and encouraging physically plausible actions through visual-dynamics constraints. With a causal design that prevents future-video tokens from influencing action tokens, explicit future-video generation is optional at inference time, allowing faster action prediction during deployment. To support this paradigm, we curate a diverse, large-scale robot dataset to pre-train an action-centered video generation model, which is then adapted as the backbone for robot policy learning. Experimental results on real-world robotic platforms show that GigaWorld-Policy runs 9x faster than the leading WAM baseline, Motus, while improving task success rates by 7%. Moreover, compared with pi-0.5, GigaWorld-Policy improves performance by 95% on RoboTwin 2.0.

Rui Cai, Jun Guo, Xinze He et al.

In this report, we introduce Xiaomi-Robotics-0, an advanced vision-language-action (VLA) model optimized for high performance and fast and smooth real-time execution. The key to our method lies in a carefully designed training recipe and deployment strategy. Xiaomi-Robotics-0 is first pre-trained on large-scale cross-embodiment robot trajectories and vision-language data, endowing it with broad and generalizable action-generation capabilities while avoiding catastrophic forgetting of the visual-semantic knowledge of the underlying pre-trained VLM. During post-training, we propose several techniques for training the VLA model for asynchronous execution to address the inference latency during real-robot rollouts. During deployment, we carefully align the timesteps of consecutive predicted action chunks to ensure continuous and seamless real-time rollouts. We evaluate Xiaomi-Robotics-0 extensively in simulation benchmarks and on two challenging real-robot tasks that require precise and dexterous bimanual manipulation. Results show that our method achieves state-of-the-art performance across all simulation benchmarks. Moreover, Xiaomi-Robotics-0 can roll out fast and smoothly on real robots using a consumer-grade GPU, achieving high success rates and throughput on both real-robot tasks. To facilitate future research, code and model checkpoints are open-sourced at https://xiaomi-robotics-0.github.io

Pan Wang, Jie Li, Jieru Chen et al.

In this paper, a convolutional sparse coding method based on global structure characteristics and spectral correlation is proposed for the reconstruction of compressive spectral images. The spectral data is regarded as the convolution sum of the convolution kernel and the corresponding coefficients, using the convolution kernel operates the global image information, preserving the structure information of the spectral image in the spatial dimension. To take full exploration of the constraints between spectra, the coefficients corresponding to the convolution kernel are constrained by the L_(2,1)norm to improve spectral accuracy. And, to solve the problem that convolutional sparse coding is insensitive to low frequency, the global total-variation (TV) constraint is added to estimate the low-frequency components. It not only ensures the effective estimation of the low-frequency but also transforms the convolutional sparse coding into a de-noising process, which makes the reconstructing process simpler. Simulations show that compared with the current mainstream optimization methods, the proposed method can improve the reconstruction quality by up to 4 dB in PSNR and 10% in SSIM, and has a great improvement in the details of the reconstructed image.

Jiawei Sun, Ruoxin Chen, Jie Li et al.

Graph Contrastive Learning (GCL) has shown promising performance in graph representation learning (GRL) without the supervision of manual annotations. GCL can generate graph-level embeddings by maximizing the Mutual Information (MI) between different augmented views of the same graph (positive pairs). However, the GCL is limited by dimensional collapse, i.e., embedding vectors only occupy a low-dimensional subspace. In this paper, we show that the smoothing effect of the graph pooling and the implicit regularization of the graph convolution are two causes of the dimensional collapse in GCL. To mitigate the above issue, we propose a non-maximum removal graph contrastive learning approach (nmrGCL), which removes "prominent'' dimensions (i.e., contribute most in similarity measurement) for positive pair in the pre-text task. Comprehensive experiments on various benchmark datasets are conducted to demonstrate the effectiveness of nmrGCL, and the results show that our model outperforms the state-of-the-art methods. Source code will be made publicly available.

Zhusi Zhong, Jie Li, Lulu Bi et al.

Medical image segmentation based on deep learning often fails when deployed on images from a different domain. The domain adaptation methods aim to solve domain-shift challenges, but still face some problems. The transfer learning methods require annotation on the target domain, and the generative unsupervised domain adaptation (UDA) models ignore domain-specific representations, whose generated quality highly restricts segmentation performance. In this study, we propose a novel Structure-Modal Constrained (SMC) UDA framework based on a discriminative paradigm and introduce edge structure as a bridge between domains. The proposed multi-modal learning backbone distills structure information from image texture to distinguish domain-invariant edge structure. With the structure-constrained self-learning and progressive ROI, our methods segment the kidney by locating the 3D spatial structure of the edge. We evaluated SMC-UDA on public renal segmentation datasets, adapting from the labeled source domain (CT) to the unlabeled target domain (CT/MRI). The experiments show that our proposed SMC-UDA has a strong generalization and outperforms generative UDA methods.

Pavel Tokmakov, Allan Jabri, Jie Li et al.

This paper proposes a self-supervised objective for learning representations that localize objects under occlusion - a property known as object permanence. A central question is the choice of learning signal in cases of total occlusion. Rather than directly supervising the locations of invisible objects, we propose a self-supervised objective that requires neither human annotation, nor assumptions about object dynamics. We show that object permanence can emerge by optimizing for temporal coherence of memory: we fit a Markov walk along a space-time graph of memories, where the states in each time step are non-Markovian features from a sequence encoder. This leads to a memory representation that stores occluded objects and predicts their motion, to better localize them. The resulting model outperforms existing approaches on several datasets of increasing complexity and realism, despite requiring minimal supervision, and hence being broadly applicable.

Jie Li, Hancheng Cao, Laura Lin et al.

Among creative professionals, Generative Artificial Intelligence (GenAI) has sparked excitement over its capabilities and fear over unanticipated consequences. How does GenAI impact User Experience Design (UXD) practice, and are fears warranted? We interviewed 20 UX Designers, with diverse experience and across companies (startups to large enterprises). We probed them to characterize their practices, and sample their attitudes, concerns, and expectations. We found that experienced designers are confident in their originality, creativity, and empathic skills, and find GenAI's role as assistive. They emphasized the unique human factors of "enjoyment" and "agency", where humans remain the arbiters of "AI alignment". However, skill degradation, job replacement, and creativity exhaustion can adversely impact junior designers. We discuss implications for human-GenAI collaboration, specifically copyright and ownership, human creativity and agency, and AI literacy and access. Through the lens of responsible and participatory AI, we contribute a deeper understanding of GenAI fears and opportunities for UXD.

Mingkang Dong, Hongyi Cai, Xiwen Lei et al.

Multimodal instruction tuning is the de facto recipe for adapting vision language models (VLMs), yet instruction data are highly redundant, making data selection critical for training efficiency. Existing methods derive selection signals from a specific model or dataset, so whenever the target model or candidate pool changes, the criteria must be recomputed from scratch at substantial cost. To address this, we propose OFA, a data selection framework that trains a reusable selector once and applies it to any dataset or model without recomputation. OFA clusters multimodal instructions in a frozen CLIP space, derives pseudo labels from the cluster structure, and trains a lightweight selector for only a few epochs; samples on which this selector is least confident are selected as the most informative. Once trained, the frozen selector transfers directly across datasets and model scales. The selector is trained once on LLaVA-665K and applied both to LLaVA-665K itself and, without any retraining, to the unseen Vision-Flan-186K. Selecting only 15% of the data, OFA achieves 98.3% of full data performance across 10 downstream benchmarks; on the smaller Vision-Flan-186K, the transferred selector surpasses full data training by 10.6%, confirming that the learned signal generalizes to datasets never seen during selector training. The same selected subsets benefit VLMs at both Qwen2.5-VL-3B and LLaVA-v1.5-7B without per model recomputation, decoupling selection from the target model. These results demonstrate that a single, transferable selector provides an effective and reusable solution for efficient multimodal instruction tuning.

Zaiyan Zhang, Jie Li, Shaowei Shi et al.

Optical remote sensing imagery is indispensable for Earth observation, yet persistent cloud occlusion limits its downstream utility. Most cloud removal (CR) methods are optimized for low-level fidelity and can over-smooth textures and boundaries that are critical for analysis-ready data (ARD), leading to a mismatch between visually plausible restoration and semantic utility. To bridge this gap, we propose TDP-CR, a task-driven multimodal framework that jointly performs cloud removal and land-cover segmentation. Central to our approach is a Prompt-Guided Fusion (PGF) mechanism, which utilizes a learnable degradation prompt to encode cloud thickness and spatial uncertainty. By combining global channel context with local prompt-conditioned spatial bias, PGF adaptively integrates Synthetic Aperture Radar (SAR) information only where optical data is corrupted. We further introduce a parameter-efficient two-phase training strategy that decouples reconstruction and semantic representation learning. Experiments on the LuojiaSET-OSFCR dataset demonstrate the superiority of our framework: TDP-CR surpasses heavy state-of-the-art baselines by 0.18 dB in PSNR while using only 15\% of the parameters, and achieves a 1.4\% improvement in mIoU consistently against multi-task competitors, effectively delivering analysis-ready data.

Chenxu Niu, Wei Zhang, Jie Li et al.

Large language model (LLM) services now answer billions of queries per day, and industry reports show that inference, not training, accounts for more than 90% of total power consumption. However, existing benchmarks focus on either training/fine-tuning or performance of inference and provide little support for power consumption measurement and analysis of inference. We introduce TokenPowerBench, the first lightweight and extensible benchmark designed for LLM-inference power consumption studies. The benchmark combines (i) a declarative configuration interface covering model choice, prompt set, and inference engine, (ii) a measurement layer that captures GPU-, node-, and system-level power without specialized power meters, and (iii) a phase-aligned metrics pipeline that attributes energy to the prefill and decode stages of every request. These elements make it straight-forward to explore the power consumed by an LLM inference run; furthermore, by varying batch size, context length, parallelism strategy and quantization, users can quickly assess how each setting affects joules per token and other energy-efficiency metrics. We evaluate TokenPowerBench on four of the most widely used model series (Llama, Falcon, Qwen, and Mistral). Our experiments cover from 1 billion parameters up to the frontier-scale Llama3-405B model. Furthermore, we release TokenPowerBench as open source to help users to measure power consumption, forecast operating expenses, and meet sustainability targets when deploying LLM services.

Jie Li, Bing Tang, Feng Wu

Real-time whole-body teleoperation is a critical method for humanoid robots to perform complex tasks in unstructured environments. However, developing a unified controller that robustly supports diverse human motions remains a significant challenge. Existing methods typically distill multiple expert policies into a single general policy, which often inevitably leads to performance degradation, particularly on highly dynamic motions. This paper presents TeleGate, a unified whole-body teleoperation framework for humanoid robots that achieves high-precision tracking across various motions while avoiding the performance loss inherent in knowledge distillation. Our key idea is to preserve the full capability of domain-specific expert policies by training a lightweight gating network, which dynamically activates experts in real-time based on proprioceptive states and reference trajectories. Furthermore, to compensate for the absence of future reference trajectories in real-time teleoperation, we introduce a VAE-based motion prior module that extracts implicit future motion intent from historical observations, enabling anticipatory control for motions requiring prediction such as jumping and standing up. We conducted empirical evaluations in simulation and also deployed our technique on the Unitree G1 humanoid robot. Using only 2.5 hours of motion capture data for training, our TeleGate achieves high-precision real-time teleoperation across diverse dynamic motions (e.g., running, fall recovery, and jumping), significantly outperforming the baseline methods in both tracking accuracy and success rate.

Chang Sun, Zili Wang, Shuyou Zhang et al.

As one of the most widely used metal tube bending methods, the rotary draw bending (RDB) process enables reliable and high-precision metal tube bending forming (MTBF). The forming accuracy is seriously affected by the springback and other potential forming defects, of which the mechanism analysis is difficult to deal with. At the same time, the existing methods are mainly conducted in offline space, ignoring the real-time information in the physical world, which is unreliable and inefficient. To address this issue, a digital-twin-enhanced (DT-enhanced) metal tube bending forming real-time prediction method based on multi-source-input multi-task learning (MTL) is proposed. The new method can achieve comprehensive MTBF real-time prediction. By sharing the common feature of the multi-close domain and adopting group regularization strategy on feature sharing and accepting layers, the accuracy and efficiency of the multi-source-input MTL can be guaranteed. Enhanced by DT, the physical real-time deformation data is aligned in the image dimension by an improved Grammy Angle Field (GAF) conversion, realizing the reflection of the actual processing. Different from the traditional offline prediction methods, the new method integrates the virtual and physical data to achieve a more efficient and accurate real-time prediction result. and the DT mapping connection between virtual and physical systems can be achieved. To exclude the effects of equipment errors, the effectiveness of the proposed method is verified on the physical experiment-verified FE simulation scenarios. At the same time, the common pre-training networks are compared with the proposed method. The results show that the proposed DT-enhanced prediction method is more accurate and efficient.

Han Xu, Zheming Zuo, Jie Li et al.

Situating at the core of Artificial Intelligence (AI), Machine Learning (ML), and more specifically, Deep Learning (DL) have embraced great success in the past two decades. However, unseen class label prediction is far less explored due to missing classes being invisible in training ML or DL models. In this work, we propose a fuzzy inference system to cope with such a challenge by adopting TSK+ fuzzy inference engine in conjunction with the Curvature-based Feature Selection (CFS) method. The practical feasibility of our system has been evaluated by predicting the positioning labels of networking devices within the realm of the Internet of Things (IoT). Competitive prediction performance confirms the efficiency and efficacy of our system, especially when a large number of continuous class labels are unseen during the model training stage.

Jingliang Duan, Jie Li, Xuyang Chen et al.

In recent times, significant advancements have been made in delving into the optimization landscape of policy gradient methods for achieving optimal control in linear time-invariant (LTI) systems. Compared with state-feedback control, output-feedback control is more prevalent since the underlying state of the system may not be fully observed in many practical settings. This paper analyzes the optimization landscape inherent to policy gradient methods when applied to static output feedback (SOF) control in discrete-time LTI systems subject to quadratic cost. We begin by establishing crucial properties of the SOF cost, encompassing coercivity, L-smoothness, and M-Lipschitz continuous Hessian. Despite the absence of convexity, we leverage these properties to derive novel findings regarding convergence (and nearly dimension-free rate) to stationary points for three policy gradient methods, including the vanilla policy gradient method, the natural policy gradient method, and the Gauss-Newton method. Moreover, we provide proof that the vanilla policy gradient method exhibits linear convergence towards local minima when initialized near such minima. The paper concludes by presenting numerical examples that validate our theoretical findings. These results not only characterize the performance of gradient descent for optimizing the SOF problem but also provide insights into the effectiveness of general policy gradient methods within the realm of reinforcement learning.

Chuantao Li, Zhi Li, Jiahao Xu et al.

Numerous studies attempt to mitigate classification bias caused by class imbalance. However, existing studies have yet to explore the collaborative optimization of imbalanced learning and model training. This constraint hinders further performance improvements. To bridge this gap, this study proposes a collaborative optimization Boosting model of multiclass imbalanced learning. This model is simple but effective by integrating the density factor and the confidence factor, this study designs a noise-resistant weight update mechanism and a dynamic sampling strategy. Rather than functioning as independent components, these modules are tightly integrated to orchestrate weight updates, sample region partitioning, and region-guided sampling. Thus, this study achieves the collaborative optimization of imbalanced learning and model training. Extensive experiments on 20 public imbalanced datasets demonstrate that the proposed model significantly outperforms eight state-of-the-art baselines. The code for the proposed model is available at: https://github.com/ChuantaoLi/DARG.

Decheng Liu, Weijie He, Chunlei Peng et al.

Face attribute evaluation plays an important role in video surveillance and face analysis. Although methods based on convolution neural networks have made great progress, they inevitably only deal with one local neighborhood with convolutions at a time. Besides, existing methods mostly regard face attribute evaluation as the individual multi-label classification task, ignoring the inherent relationship between semantic attributes and face identity information. In this paper, we propose a novel \textbf{trans}former-based representation for \textbf{f}ace \textbf{a}ttribute evaluation method (\textbf{TransFA}), which could effectively enhance the attribute discriminative representation learning in the context of attention mechanism. The multiple branches transformer is employed to explore the inter-correlation between different attributes in similar semantic regions for attribute feature learning. Specially, the hierarchical identity-constraint attribute loss is designed to train the end-to-end architecture, which could further integrate face identity discriminative information to boost performance. Experimental results on multiple face attribute benchmarks demonstrate that the proposed TransFA achieves superior performances compared with state-of-the-art methods.

Jie Li, Zhixin Li, Zhi Liu et al.

Volumetric video, also known as hologram video, is a novel medium that portrays natural content in Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR). It is expected to be the next-gen video technology and a prevalent use case for 5G and beyond wireless communication. Considering that each user typically only watches a section of the volumetric video, known as the viewport, it is essential to have precise viewport prediction for optimal performance. However, research on this topic is still in its infancy. In the end, this paper presents and proposes a novel approach, named Saliency and Trajectory Viewport Prediction (STVP), which aims to improve the precision of viewport prediction in volumetric video streaming. The STVP extensively utilizes video saliency information and viewport trajectory. To our knowledge, this is the first comprehensive study of viewport prediction in volumetric video streaming. In particular, we introduce a novel sampling method, Uniform Random Sampling (URS), to reduce computational complexity while still preserving video features in an efficient manner. Then we present a saliency detection technique that incorporates both spatial and temporal information for detecting static, dynamic geometric, and color salient regions. Finally, we intelligently fuse saliency and trajectory information to achieve more accurate viewport prediction. We conduct extensive simulations to evaluate the effectiveness of our proposed viewport prediction methods using state-of-the-art volumetric video sequences. The experimental results show the superiority of the proposed method over existing schemes. The dataset and source code will be publicly accessible after acceptance.

Xiangrui Cai, Shaocheng Ma, Lei Cao et al.

Electroencephalography (EEG) signal decoding is a key technology that translates brain activity into executable commands, laying the foundation for direct brain-machine interfacing and intelligent interaction. To address the inherent spatiotemporal heterogeneity of EEG signals, this paper proposes a multi-branch parallel architecture, where each temporal scale is equipped with an independent spatial feature extraction module. To further enhance multi-branch feature fusion, we propose a Fusion of Multiscale Features via Centralized Sparse-attention Network (EEG-CSANet), a centralized sparse-attention network. It employs a main-auxiliary branch architecture, where the main branch models core spatiotemporal patterns via multiscale self-attention, and the auxiliary branch facilitates efficient local interactions through sparse cross-attention. Experimental results show that EEG-CSANet achieves state-of-the-art (SOTA) performance across five public datasets (BCIC-IV-2A, BCIC-IV-2B, HGD, SEED, and SEED-VIG), with accuracies of 88.54%, 91.09%, 99.43%, 96.03%, and 90.56%, respectively. Such performance demonstrates its strong adaptability and robustness across various EEG decoding tasks. Moreover, extensive ablation studies are conducted to enhance the interpretability of EEG-CSANet. In the future, we hope that EEG-CSANet could serve as a promising baseline model in the field of EEG signal decoding. The source code is publicly available at: https://github.com/Xiangrui-Cai/EEG-CSANet

Shijing Yuan, Hongze Liu, Hongtao Lv et al.

Cross-silo federated learning (FL) is a typical FL that enables organizations(e.g., financial or medical entities) to train global models on isolated data. Reasonable incentive is key to encouraging organizations to contribute data. However, existing works on incentivizing cross-silo FL lack consideration of the environmental dynamics (e.g., precision of the trained global model and data owned by uncertain clients during the training processes). Moreover, most of them assume that organizations share private information, which is unrealistic. To overcome these limitations, we propose a novel adaptive mechanism for cross-silo FL, towards incentivizing organizations to contribute data to maximize their long-term payoffs in a real dynamic training environment. The mechanism is based on multi-agent reinforcement learning, which learns near-optimal data contribution strategy from the history of potential games without organizations' private information. Experiments demonstrate that our mechanism achieves adaptive incentive and effectively improves the long-term payoffs for organizations.

Bowen Li, Hanlin Gu, Ruoxin Chen et al.

Federated Learning (FL) has emerged as a promising approach for collaborative model training without sharing private data. However, privacy concerns regarding information exchanged during FL have received significant research attention. Gradient Inversion Attacks (GIAs) have been proposed to reconstruct the private data retained by local clients from the exchanged gradients. While recovering private data, the data dimensions and the model complexity increase, which thwart data reconstruction by GIAs. Existing methods adopt prior knowledge about private data to overcome those challenges. In this paper, we first observe that GIAs with gradients from a single iteration fail to reconstruct private data due to insufficient dimensions of leaked gradients, complex model architectures, and invalid gradient information. We investigate a Temporal Gradient Inversion Attack with a Robust Optimization framework, called TGIAs-RO, which recovers private data without any prior knowledge by leveraging multiple temporal gradients. To eliminate the negative impacts of outliers, e.g., invalid gradients for collaborative optimization, robust statistics are proposed. Theoretical guarantees on the recovery performance and robustness of TGIAs-RO against invalid gradients are also provided. Extensive empirical results on MNIST, CIFAR10, ImageNet and Reuters 21578 datasets show that the proposed TGIAs-RO with 10 temporal gradients improves reconstruction performance compared to state-of-the-art methods, even for large batch sizes (up to 128), complex models like ResNet18, and large datasets like ImageNet (224*224 pixels). Furthermore, the proposed attack method inspires further exploration of privacy-preserving methods in the context of FL.

Hongbo Jiang, Jie Li, Xinqi Cai et al.

Practical cloud-edge deployment of Cross-Modal Re-identification (CM-ReID) faces challenges due to maintaining a fragmented ecosystem of specialized cloud models for diverse modalities. While Multi-Modal Large Language Models (MLLMs) offer strong unification potential, existing approaches fail to adapt them into a single end-to-end backbone and lack effective knowledge distillation strategies for edge deployment. To address these limitations, we propose MLLMEmbed-ReID, a unified framework based on a powerful cloud-edge architecture. First, we adapt a foundational MLLM into a state-of-the-art cloud model. We leverage instruction-based prompting to guide the MLLM in generating a unified embedding space across RGB, infrared, sketch, and text modalities. This model is then trained efficiently with a hierarchical Low-Rank Adaptation finetuning (LoRA-SFT) strategy, optimized under a holistic cross-modal alignment objective. Second, to deploy its knowledge onto an edge-native student, we introduce a novel distillation strategy motivated by the low-rank property in the teacher's feature space. To prioritize essential information, this method employs a Principal Component Mapping loss, while relational structures are preserved via a Feature Relation loss. Our lightweight edge-based model achieves state-of-the-art performance on multiple visual CM-ReID benchmarks, while its cloud-based counterpart excels across all CM-ReID benchmarks. The MLLMEmbed-ReID framework thus presents a complete and effective solution for deploying unified MLLM-level intelligence on resource-constrained devices. The code and models will be open-sourced soon.

Ye Bai, Jie Li, Wenjing Han et al.

While transformers and their variant conformers show promising performance in speech recognition, the parameterized property leads to much memory cost during training and inference. Some works use cross-layer weight-sharing to reduce the parameters of the model. However, the inevitable loss of capacity harms the model performance. To address this issue, this paper proposes a parameter-efficient conformer via sharing sparsely-gated experts. Specifically, we use sparsely-gated mixture-of-experts (MoE) to extend the capacity of a conformer block without increasing computation. Then, the parameters of the grouped conformer blocks are shared so that the number of parameters is reduced. Next, to ensure the shared blocks with the flexibility of adapting representations at different levels, we design the MoE routers and normalization individually. Moreover, we use knowledge distillation to further improve the performance. Experimental results show that the proposed model achieves competitive performance with 1/3 of the parameters of the encoder, compared with the full-parameter model.

Jingliang Duan, Jie Li, Yinsong Ma et al.

Understanding the optimization landscape of linear quadratic regulation (LQR) problems is fundamental to the design of efficient reinforcement learning solutions. Recent work has made significant progress in characterizing the landscape of static output-feedback control and linear quadratic Gaussian (LQG) control. For LQG, much of the analysis leverages the separation principle, which allows the controller and estimator to be designed independently. However, this simplification breaks down when the gradients with respect to the estimator and controller parameters are inherently coupled, leading to a more intricate analysis. This paper investigates the optimization landscape of observer-based dynamic output-feedback control of LQR problems. We derive the optimal observer-controller pair in settings where transient quadratic performance cannot be neglected. Our analysis reveals that, in general, the combination of the standard LQR controller and the observer that minimizes the trace of the accumulated estimation error covariance does not correspond to a stationary point of the overall closed-loop performance objective. Moreover, we derive a pair of discrete-time Sylvester equations with symmetric structure, both involving the same set of matrix elements, that characterize the stationary point of the observer-based dynamic LQR problem. These equations offer analytical insight into the structure of the optimality conditions and provide a foundation for developing numerical policy gradient methods aimed at learning complex controllers that rely on reconstructed state information.

Chang Liu, Juan Li, Sheng Zhang et al.

Depth estimation is one of the key technologies for realizing 3D perception in unmanned systems. Monocular depth estimation has been widely researched because of its low-cost advantage, but the existing methods face the challenges of poor depth estimation performance and blurred object boundaries on embedded systems. In this paper, we propose a novel monocular depth estimation model, BoRe-Depth, which contains only 8.7M parameters. It can accurately estimate depth maps on embedded systems and significantly improves boundary quality. Firstly, we design an Enhanced Feature Adaptive Fusion Module (EFAF) which adaptively fuses depth features to enhance boundary detail representation. Secondly, we integrate semantic knowledge into the encoder to improve the object recognition and boundary perception capabilities. Finally, BoRe-Depth is deployed on NVIDIA Jetson Orin, and runs efficiently at 50.7 FPS. We demonstrate that the proposed model significantly outperforms previous lightweight models on multiple challenging datasets, and we provide detailed ablation studies for the proposed methods. The code is available at https://github.com/liangxiansheng093/BoRe-Depth.