Yujia Zhang

Yuzhi Zhao, Lai-Man Po, Xuehui Wang et al.

The appearances of children are inherited from their parents, which makes it feasible to predict them. Predicting realistic children's faces may help settle many social problems, such as age-invariant face recognition, kinship verification, and missing child identification. It can be regarded as an image-to-image translation task. Existing approaches usually assume domain information in the image-to-image translation can be interpreted by "style", i.e., the separation of image content and style. However, such separation is improper for the child face prediction, because the facial contours between children and parents are not the same. To address this issue, we propose a new disentangled learning strategy for children's face prediction. We assume that children's faces are determined by genetic factors (compact family features, e.g., face contour), external factors (facial attributes irrelevant to prediction, such as moustaches and glasses), and variety factors (individual properties for each child). On this basis, we formulate predictions as a mapping from parents' genetic factors to children's genetic factors, and disentangle them from external and variety factors. In order to obtain accurate genetic factors and perform the mapping, we propose a ChildPredictor framework. It transfers human faces to genetic factors by encoders and back by generators. Then, it learns the relationship between the genetic factors of parents and children through a mapping function. To ensure the generated faces are realistic, we collect a large Family Face Database to train ChildPredictor and evaluate it on the FF-Database validation set. Experimental results demonstrate that ChildPredictor is superior to other well-known image-to-image translation methods in predicting realistic and diverse child faces. Implementation codes can be found at https://github.com/zhaoyuzhi/ChildPredictor.

Yuzhi Zhao, Lai-Man Po, Kangcheng Liu et al.

In this paper, we propose a scribble-based video colorization network with temporal aggregation called SVCNet. It can colorize monochrome videos based on different user-given color scribbles. It addresses three common issues in the scribble-based video colorization area: colorization vividness, temporal consistency, and color bleeding. To improve the colorization quality and strengthen the temporal consistency, we adopt two sequential sub-networks in SVCNet for precise colorization and temporal smoothing, respectively. The first stage includes a pyramid feature encoder to incorporate color scribbles with a grayscale frame, and a semantic feature encoder to extract semantics. The second stage finetunes the output from the first stage by aggregating the information of neighboring colorized frames (as short-range connections) and the first colorized frame (as a long-range connection). To alleviate the color bleeding artifacts, we learn video colorization and segmentation simultaneously. Furthermore, we set the majority of operations on a fixed small image resolution and use a Super-resolution Module at the tail of SVCNet to recover original sizes. It allows the SVCNet to fit different image resolutions at the inference. Finally, we evaluate the proposed SVCNet on DAVIS and Videvo benchmarks. The experimental results demonstrate that SVCNet produces both higher-quality and more temporally consistent videos than other well-known video colorization approaches. The codes and models can be found at https://github.com/zhaoyuzhi/SVCNet.

Xianzhe Fan, Shengliang Deng, Xiaoyang Wu et al.

Existing Vision-Language-Action (VLA) models typically take 2D images as visual input, which limits their spatial understanding in complex scenes. How can we incorporate 3D information to enhance VLA capabilities? We conduct a pilot study across different observation spaces and visual representations. The results show that explicitly lifting visual input into point clouds yields representations that better complement their corresponding 2D representations. To address the challenges of (1) scarce 3D data and (2) the domain gap induced by cross-environment differences and depth-scale biases, we propose Any3D-VLA. It unifies the simulator, sensor, and model-estimated point clouds within a training pipeline, constructs diverse inputs, and learns domain-agnostic 3D representations that are fused with the corresponding 2D representations. Simulation and real-world experiments demonstrate Any3D-VLA's advantages in improving performance and mitigating the domain gap. Our project homepage is available at https://xianzhefan.github.io/Any3D-VLA.github.io.

Zhenhua Xu, Yujia Zhang, Enze Xie et al.

Multimodal large language models (MLLMs) have emerged as a prominent area of interest within the research community, given their proficiency in handling and reasoning with non-textual data, including images and videos. This study seeks to extend the application of MLLMs to the realm of autonomous driving by introducing DriveGPT4, a novel interpretable end-to-end autonomous driving system based on LLMs. Capable of processing multi-frame video inputs and textual queries, DriveGPT4 facilitates the interpretation of vehicle actions, offers pertinent reasoning, and effectively addresses a diverse range of questions posed by users. Furthermore, DriveGPT4 predicts low-level vehicle control signals in an end-to-end fashion.These advanced capabilities are achieved through the utilization of a bespoke visual instruction tuning dataset, specifically tailored for autonomous driving applications, in conjunction with a mix-finetuning training strategy. DriveGPT4 represents the pioneering effort to leverage LLMs for the development of an interpretable end-to-end autonomous driving solution. Evaluations conducted on the BDD-X dataset showcase the superior qualitative and quantitative performance of DriveGPT4. Additionally, the fine-tuning of domain-specific data enables DriveGPT4 to yield close or even improved results in terms of autonomous driving grounding when contrasted with GPT4-V.

Zhiwen Jing, Ziliang Zhao, Yang Feng et al.

Supply Chain Platforms (SCPs) provide downstream industries with numerous raw materials. Compared with traditional e-commerce platforms, data in SCPs is more sparse due to limited user interests. To tackle the data sparsity problem, one can apply Cross-Domain Recommendation (CDR) which improves the recommendation performance of the target domain with the source domain information. However, applying CDR to SCPs directly ignores the hierarchical structure of commodities in SCPs, which reduce the recommendation performance. To leverage this feature, in this paper, we take the catering platform as an example and propose GReS, a graphical cross-domain recommendation model. The model first constructs a tree-shaped graph to represent the hierarchy of different nodes of dishes and ingredients, and then applies our proposed Tree2vec method combining GCN and BERT models to embed the graph for recommendations. Experimental results on a commercial dataset show that GReS significantly outperforms state-of-the-art methods in Cross-Domain Recommendation for Supply Chain Platforms.

Jian Li, Haojing Huang, Yujia Zhang et al.

Recently, there has been significant interest in replacing the reward model in Reinforcement Learning with Human Feedback (RLHF) methods for Large Language Models (LLMs), such as Direct Preference Optimization (DPO) and its variants. These approaches commonly use a binary cross-entropy mechanism on pairwise samples, i.e., minimizing and maximizing the loss based on preferred or dis-preferred responses, respectively. However, while this training strategy omits the reward model, it also overlooks the varying preference degrees within different responses. We hypothesize that this is a key factor hindering LLMs from sufficiently understanding human preferences. To address this problem, we propose a novel Self-supervised Preference Optimization (SPO) framework, which constructs a self-supervised preference degree loss combined with the alignment loss, thereby helping LLMs improve their ability to understand the degree of preference. Extensive experiments are conducted on two widely used datasets of different tasks. The results demonstrate that SPO can be seamlessly integrated with existing preference optimization methods and significantly boost their performance to achieve state-of-the-art performance. We also conduct detailed analyses to offer comprehensive insights into SPO, which verifies its effectiveness. The code is available at https://github.com/lijian16/SPO.

Mostafa Ahmadi, Amin Alizadeh Naeini, Mohammad Moein Sheikholeslami et al.

This paper proposes a novel visual simultaneous localization and mapping (SLAM) system called Hybrid Depth-augmented Panoramic Visual SLAM (HDPV-SLAM), that employs a panoramic camera and a tilted multi-beam LiDAR scanner to generate accurate and metrically-scaled trajectories. RGB-D SLAM was the design basis for HDPV-SLAM, which added depth information to visual features. It aims to solve the two major issues hindering the performance of similar SLAM systems. The first obstacle is the sparseness of LiDAR depth, which makes it difficult to correlate it with the extracted visual features of the RGB image. A deep learning-based depth estimation module for iteratively densifying sparse LiDAR depth was suggested to address this issue. The second issue pertains to the difficulties in depth association caused by a lack of horizontal overlap between the panoramic camera and the tilted LiDAR sensor. To surmount this difficulty, we present a hybrid depth association module that optimally combines depth information estimated by two independent procedures, feature-based triangulation and depth estimation. During a phase of feature tracking, this hybrid depth association module aims to maximize the use of more accurate depth information between the triangulated depth with visual features tracked and the deep learning-based corrected depth. We evaluated the efficacy of HDPV-SLAM using the 18.95 km-long York University and Teledyne Optech (YUTO) MMS dataset. The experimental results demonstrate that the two proposed modules contribute substantially to the performance of HDPV-SLAM, which surpasses that of the state-of-the-art (SOTA) SLAM systems.

Songlin Yang, Haobin Zhong, Ruilin Zhang et al.

The rapid evolution of generative video foundation models has propelled the field toward professional-grade cinematic synthesis. To achieve such demanding quality, the community transitions towards Reinforcement Learning (RL) and agentic workflows. However, reliable evaluation has emerged as a critical bottleneck. Existing benchmarks predominantly evaluate ''whether it is right'' (basic prompt-following) while fundamentally neglecting ''whether it is good'' (cinematic quality, acting, and aesthetics). Furthermore, current automated metrics lack the domain-specific rigor required to provide trustworthy signals, creating a severe credibility gap between human aesthetic perception and machine scoring. To bridge this gap, we introduce EvalVerse, a comprehensive, pipeline-aware, and expert-calibrated evaluation framework. We treat video generation assessment not merely as an engineering task, but as a core scientific problem: the systematic digitization of subjective cinematic expertise. First, we organize domain knowledge into an evaluation taxonomy aligned with the professional filmmaking workflow (pre-production, production, and post-production). Second, we distill human expert judgments into a curated dataset with large-scale human annotations. Third, we inject this knowledge into Vision-Language Models (VLMs) through an expert-calibrated fine-tuning strategy, enabling the VLM to perform explicit Chain-of-Thought reasoning. Compared to previous works, EvalVerse not only retains compatibility with foundational ''rightness'' metrics, but also significantly expands the criteria to ''goodness'' and broaden the task coverage to complex multi-shot sequencing and audio-visual integration. Consequently, by providing granular diagnostic signals, EvalVerse transcends a static leaderboard and establishes a fundamental infrastructure for future work, such as reward models and evaluator agent.

Dazhao Du, Jian Liu, Jialong Qin et al.

Video large language models (Video LLMs) achieve strong benchmark accuracy, yet often answer video questions through shortcuts such as single-frame cues and language priors rather than by tracking spatiotemporal dynamics. This issue is exacerbated in RL post-training, where correctness-only rewards can further reinforce shortcut policies that obtain high reward without tracking video dynamics. We address this by asking a controlled counterfactual question: if the visual world changed while the question remained fixed, should the answer change or stay the same? Based on this view, we propose \textbf{Counterfactual Relational Policy Optimization (CRPO)}, a dual-branch RL framework for improving \emph{spatiotemporal sensitivity}. CRPO constructs counterfactual videos through horizontal flips and temporal reversals, trains on both original and counterfactual branches, and introduces a \textbf{Counterfactual Relation Reward (CRR)} between their answers. CRR encourages answers to change for dynamic questions and remain unchanged for static questions. This cross-branch constraint makes it difficult for shortcut policies to be consistently rewarded across both branches. To evaluate this property, we introduce \textbf{DyBench}, a paired counterfactual video benchmark with 3,014 videos covering reversible dynamics, moving direction, and event sequence, together with a strict pair-accuracy metric that prevents fixed-answer shortcuts from inflating scores. Experiments show that CRPO outperforms prior RL methods on spatiotemporal-sensitive evaluations while maintaining competitive general video performance. On Qwen3-VL-8B, CRPO improves DyBench P-Acc by +7.7 and TimeBlind I-Acc by +8.2 over the base model, indicating improved spatiotemporal sensitivity rather than stronger reliance on static shortcuts. The project website can be found at https://ddz16.github.io/crpo.github.io/ .

Dazhao Du, Liao Duan, Jian Liu et al.

Video temporal grounding (VTG), which localizes the start and end times of a queried event in an untrimmed video, is a key test of whether multimodal large language models (MLLMs) understand not only what happens but also when it happens. Although modern MLLMs describe video content fluently, their timestamp predictions remain unreliable, while existing remedies either require costly post-training on temporal annotations or rely on coarse training-free heuristics. In this work, we probe the cross-modal attention of MLLMs and uncover a perception-generation gap. Our key finding is that MLLMs often know the target interval during prefill, but lose this signal when generating the final answer. In the prefill stage, a sparse set of attention heads, which we call \emph{Temporal Grounding Heads} (TG-Heads), concentrates query-to-video attention on the ground-truth interval. During autoregressive decoding, however, the answer tokens shift attention away from this interval toward visually salient but query-irrelevant segments. This observation motivates an inference-time read-then-regenerate framework. We first convert TG-Head prefill attention into a debiased frame-level relevance signal and extract the high-attention interval it highlights. We then re-invoke the MLLM with visual context restricted to this interval, using video cropping or attention masking to suppress distractors. Without parameter updates and architectural changes, our framework consistently improves MiMo-VL-7B, Qwen3-VL-8B, and TimeLens-8B on three VTG benchmarks, with gains of up to +3.5 mIoU. The project website can be found at https://ddz16.github.io/mllmsknowwhen.github.io/.

Yujia Zhang, Xiaoyang Wu, Yunhan Yang et al.

We dream of a future where point clouds from all domains can come together to shape a single model that benefits them all. Toward this goal, we present Utonia, a first step toward training a single self-supervised point transformer encoder across diverse domains, spanning remote sensing, outdoor LiDAR, indoor RGB-D sequences, object-centric CAD models, and point clouds lifted from RGB-only videos. Despite their distinct sensing geometries, densities, and priors, Utonia learns a consistent representation space that transfers across domains. This unification improves perception capability while revealing intriguing emergent behaviors that arise only when domains are trained jointly. Beyond perception, we observe that Utonia representations can also benefit embodied and multimodal reasoning: conditioning vision-language-action policies on Utonia features improves robotic manipulation, and integrating them into vision-language models yields gains on spatial reasoning. We hope Utonia can serve as a step toward foundation models for sparse 3D data, and support downstream applications in AR/VR, robotics, and autonomous driving.

Yiming Lei, Yiqi Wang, Yujia Zhang et al.

Workbook-scale spreadsheet understanding is increasingly important for language-model-based data analysis agents, but remains challenging because relevant information is often distributed across multiple sheets with heterogeneous schemas, layouts, and implicit relationships. Existing retrieval-augmented approaches typically decompose spreadsheets into rows, columns, or blocks to improve scalability; however, such chunk-centric representations can fragment worksheets into isolated text spans and weaken global sheet-level semantics. We propose Sheet as Token, a graph-enhanced framework that treats each worksheet as a unified semantic unit for multi-sheet spreadsheet retrieval. Our method extracts schema-aware records from sheet names, column headers, representative values, and layout features, and encodes each worksheet into a compact dense token. Given a natural-language query, a Graph Retriever constructs a query-specific candidate graph over sheet tokens using semantic, query-conditioned, schema-consistency, and shape-compatibility relations, and composes these channels through a multi-stage graph transformer to retrieve supporting sheet sets. Experiments on a constructed multi-sheet spreadsheet corpus show that sheet-level tokenization learns stable representations, and that graph-enhanced cross-sheet reasoning improves listwise retrieval over a shallow graph baseline with limited additional graph-side computation. These results suggest that sheet-level tokenization is a promising abstraction for scalable multi-sheet spreadsheet understanding.

Changxiao Ma, Chao Yuan, Xincheng Shi et al.

Person re-identification (ReID) suffers from a lack of large-scale high-quality training data due to challenges in data privacy and annotation costs. While previous approaches have explored pedestrian generation for data augmentation, they often fail to ensure identity consistency and suffer from insufficient controllability, thereby limiting their effectiveness in dataset augmentation. To address this, We introduce OmniPerson, the first unified identity-preserving pedestrian generation pipeline for visible/infrared image/video ReID tasks. Our contributions are threefold: 1) We proposed OmniPerson, a unified generation model, offering holistic and fine-grained control over all key pedestrian attributes. Supporting RGB/IR modality image/video generation with any number of reference images, two kinds of person poses, and text. Also including RGB-to-IR transfer and image super-resolution abilities.2) We designed Multi-Refer Fuser for robust identity preservation with any number of reference images as input, making OmniPerson could distill a unified identity from a set of multi-view reference images, ensuring our generated pedestrians achieve high-fidelity pedestrian generation.3) We introduce PersonSyn, the first large-scale dataset for multi-reference, controllable pedestrian generation, and present its automated curation pipeline which transforms public, ID-only ReID benchmarks into a richly annotated resource with the dense, multi-modal supervision required for this task. Experimental results demonstrate that OmniPerson achieves SoTA in pedestrian generation, excelling in both visual fidelity and identity consistency. Furthermore, augmenting existing datasets with our generated data consistently improves the performance of ReID models. We will open-source the full codebase, pretrained model, and the PersonSyn dataset.

Yi Pan, Yujia Zhang, Michael Kampffmeyer et al.

Partially Relevant Video Retrieval (PRVR) is a practical yet challenging task that involves retrieving videos based on queries relevant to only specific segments. While existing works follow the paradigm of developing models to process unimodal features, powerful pretrained vision-language models like CLIP remain underexplored in this field. To bridge this gap, we propose ProPy, a model with systematic architectural adaption of CLIP specifically designed for PRVR. Drawing insights from the semantic relevance of multi-granularity events, ProPy introduces two key innovations: (1) A Prompt Pyramid structure that organizes event prompts to capture semantics at multiple granularity levels, and (2) An Ancestor-Descendant Interaction Mechanism built on the pyramid that enables dynamic semantic interaction among events. With these designs, ProPy achieves SOTA performance on three public datasets, outperforming previous models by significant margins. Code is available at https://github.com/BUAAPY/ProPy.

Yujia Zhang, Lai-Man Po, Xuyuan Xu et al.

Spatio-temporal representation learning is critical for video self-supervised representation. Recent approaches mainly use contrastive learning and pretext tasks. However, these approaches learn representation by discriminating sampled instances via feature similarity in the latent space while ignoring the intermediate state of the learned representations, which limits the overall performance. In this work, taking into account the degree of similarity of sampled instances as the intermediate state, we propose a novel pretext task - spatio-temporal overlap rate (STOR) prediction. It stems from the observation that humans are capable of discriminating the overlap rates of videos in space and time. This task encourages the model to discriminate the STOR of two generated samples to learn the representations. Moreover, we employ a joint optimization combining pretext tasks with contrastive learning to further enhance the spatio-temporal representation learning. We also study the mutual influence of each component in the proposed scheme. Extensive experiments demonstrate that our proposed STOR task can favor both contrastive learning and pretext tasks. The joint optimization scheme can significantly improve the spatio-temporal representation in video understanding. The code is available at https://github.com/Katou2/CSTP.

Jianzhu Huai, Yujia Zhang, Alper Yilmaz

In recent years, commodity mobile devices equipped with cameras and inertial measurement units (IMUs) have attracted much research and design effort for augmented reality (AR) and robotics applications. Based on such sensors, many commercial AR toolkits and public benchmark datasets have been made available to accelerate hatching and validating new ideas. To lower the difficulty and enhance the flexibility in accessing the rich raw data of typical AR sensors on mobile devices, this paper present the mobile AR sensor (MARS) logger for two of the most popular mobile operating systems, Android and iOS. The logger highlights the best possible synchronization between the camera and the IMU allowed by a mobile device, and efficient saving of images at about 30Hz, and recording the metadata relevant to AR applications. This logger has been tested on a relatively large spectrum of mobile devices, and the collected data has been used for analyzing the sensor characteristics. We see that this application will facilitate research and development related to AR and robotics, so it has been open sourced at https://github.com/OSUPCVLab/mobile-ar-sensor-logger.

Yujie Feng, Jian Li, Zhihan Zhou et al.

Large Language Models (LLMs) achieve impressive performance across many tasks but remain prone to hallucination, especially in long-form generation where redundant retrieved contexts and lengthy reasoning chains amplify factual errors. Recent studies highlight a critical phenomenon: the closer key information appears to the model outputs, the higher the factual accuracy. However, existing retrieval-augmented language models (RALMs) lack effective mechanisms to ensure this proximity - external evidence is injected into reasoning via multi-turn retrieval, but this cannot ensure key information stays close to the outputs. We propose Micro-Macro Retrieval (M2R), a novel retrieve-while-generate framework to fill this gap. At the macro level, M2R retrieves coarse-grained evidence from external sources; at the micro level, it extracts essential results from a key information repository built during reasoning and reuses them while generating answers. This design directly addresses the key-information-to-output proximity bottleneck, effectively reducing hallucination in long-form tasks. M2R is trained with a curriculum learning-based reinforcement learning strategy using customized rule-based rewards, enabling stable acquisition of retrieval and grounding skills. Extensive experiments across different benchmarks demonstrate the effectiveness of M2R, especially in lengthy-context settings.

Zack Dewis, Apratim Sen, Jeffrey Wong et al.

This paper utilizes statistical data from various open datasets in Calgary to to uncover patterns and insights for community crimes, disorders, and traffic incidents. Community attributes like demographics, housing, and pet registration were collected and analyzed through geospatial visualization and correlation analysis. Strongly correlated features were identified using the chi-square test, and predictive models were built using association rule mining and machine learning algorithms. The findings suggest that crime rates are closely linked to factors such as population density, while pet registration has a smaller impact. This study offers valuable insights for city managers to enhance community safety strategies.

Ziyu Chen, Fan Zhu, Hui Zhu et al.

Recent 3D Gaussian Splatting (3DGS) methods have demonstrated the feasibility of self-driving scene reconstruction and novel view synthesis. However, most existing methods either rely solely on cameras or use LiDAR only for Gaussian initialization or depth supervision, while the rich scene information contained in point clouds, such as reflectance, and the complementarity between LiDAR and RGB have not been fully exploited, leading to degradation in challenging self-driving scenes, such as those with high ego-motion and complex lighting. To address these issues, we propose a robust and efficient LiDAR-reflectance-guided Salient Gaussian Splatting method (LR-SGS) for self-driving scenes, which introduces a structure-aware Salient Gaussian representation, initialized from geometric and reflectance feature points extracted from LiDAR and refined through a salient transform and improved density control to capture edge and planar structures. Furthermore, we calibrate LiDAR intensity into reflectance and attach it to each Gaussian as a lighting-invariant material channel, jointly aligned with RGB to enforce boundary consistency. Extensive experiments on the Waymo Open Dataset demonstrate that LR-SGS achieves superior reconstruction performance with fewer Gaussians and shorter training time. In particular, on Complex Lighting scenes, our method surpasses OmniRe by 1.18 dB PSNR.

Jian Li, Shenglin Yin, Yujia Zhang et al.

Direct Preference Optimization (DPO) is a widely used reinforcement learning from human feedback (RLHF) method across various domains. Recent research has increasingly focused on the role of token importance in improving DPO effectiveness. It is observed that identical or semantically similar content (defined as ambiguous content) frequently appears within the preference pairs. We hypothesize that the presence of ambiguous content during DPO training may introduce ambiguity, thereby limiting further improvements in alignment. Through mathematical analysis and proof-of-concept experiments, we reveal that ambiguous content may potentially introduce ambiguities, thereby degrading performance. To address this issue, we introduce Ambiguity Awareness Optimization (AAO), a simple yet effective approach that automatically re-weights ambiguous content to reduce ambiguities by calculating semantic similarity from preference pairs. Through extensive experiments, we demonstrate that AAO consistently and significantly surpasses state-of-the-art approaches in performance, without markedly increasing response length, across multiple model scales and widely adopted benchmark datasets, including AlpacaEval 2, MT-Bench, and Arena-Hard. Specifically, AAO outperforms DPO by up to 8.9 points on AlpacaEval 2 and achieves an improvement of by up to 15.0 points on Arena-Hard.

Yujia Zhang, Xiaoyang Wu, Yixing Lao et al.

Humans learn abstract concepts through multisensory synergy, and once formed, such representations can often be recalled from a single modality. Inspired by this principle, we introduce Concerto, a minimalist simulation of human concept learning for spatial cognition, combining 3D intra-modal self-distillation with 2D-3D cross-modal joint embedding. Despite its simplicity, Concerto learns more coherent and informative spatial features, as demonstrated by zero-shot visualizations. It outperforms both standalone SOTA 2D and 3D self-supervised models by 14.2% and 4.8%, respectively, as well as their feature concatenation, in linear probing for 3D scene perception. With full fine-tuning, Concerto sets new SOTA results across multiple scene understanding benchmarks (e.g., 80.7% mIoU on ScanNet). We further present a variant of Concerto tailored for video-lifted point cloud spatial understanding, and a translator that linearly projects Concerto representations into CLIP's language space, enabling open-world perception. These results highlight that Concerto emerges spatial representations with superior fine-grained geometric and semantic consistency.

Yujie Feng, Jian Li, Xiaoyu Dong et al.

Continual learning (CL) is essential for deploying large language models (LLMs) in dynamic real-world environments without the need for costly retraining. Recent model merging-based methods have attracted significant attention, but they still struggle to effectively manage the trade-off between learning new knowledge and preventing forgetting, a challenge largely stemming from suboptimal number of merges and merging frequency. In this paper, we introduce Adaptive Iterative Model Merging (AimMerging), a novel CL framework that utilizes learning and forgetting signals from the training trajectory to dynamically monitor the model's training status. Guided by dynamic monitoring, the training trajectory-guided merge controller adaptively determines the timing and frequency of iterative fusion, while the rehearsal-based knowledge fusion module computes the merging weights and executes the fusion. Comprehensive experiments on three CL benchmarks with various model sizes (from 770M to 13B) demonstrate that AimMerging achieves significant performance improvements over existing state-of-the-art methods, with an average relative improvement of 80% and 59% on FWT and BWT, respectively. The source code is provided for reproducibility.

Shuqing Li, Cuiyun Gao, Jianping Zhang et al.

The quality of Virtual Reality (VR) apps is vital, particularly the rendering quality of the VR Graphical User Interface (GUI). Different from traditional 2D apps, VR apps create a 3D digital scene for users, by rendering two distinct 2D images for the user's left and right eyes, respectively. Stereoscopic visual inconsistency (denoted as "SVI") issues, however, undermine the rendering process of the user's brain, leading to user discomfort and even adverse health effects. Such issues commonly exist but remain underexplored. We conduct an empirical analysis on 282 SVI bug reports from 15 VR platforms, summarizing 15 types of manifestations. The empirical analysis reveals that automatically detecting SVI issues is challenging, mainly because: (1) lack of training data; (2) the manifestations of SVI issues are diverse, complicated, and often application-specific; (3) most accessible VR apps are closed-source commercial software. Existing pattern-based supervised classification approaches may be inapplicable or ineffective in detecting the SVI issues. To counter these challenges, we propose an unsupervised black-box testing framework named StereoID to identify the stereoscopic visual inconsistencies, based only on the rendered GUI states. StereoID generates a synthetic right-eye image based on the actual left-eye image and computes distances between the synthetic right-eye image and the actual right-eye image to detect SVI issues. We propose a depth-aware conditional stereo image translator to power the image generation process, which captures the expected perspective shifts between left-eye and right-eye images. We build a large-scale unlabeled VR stereo screenshot dataset with larger than 171K images from 288 real-world VR apps for experiments. After substantial experiments, StereoID demonstrates superior performance for detecting SVI issues in both user reports and wild VR apps.

Alexander Ruch, Yujia Zhang, Michael Macy

Lifestyle politics emerge when activities that have no substantive relevance to ideology become politically aligned and polarized. Homophily and social influence are able generate these fault lines on their own; however, social identities from demographics may serve as coordinating mechanisms through which lifestyle politics are mobilized are spread. Using a dataset of 137,661,886 observations from 299,327 Facebook interests aggregated across users of different racial/ethnic, education, age, gender, and income demographics, we find that the most extreme instances of lifestyle politics are those which are highly confounded by demographics such as race/ethnicity (e.g., Black artists and performers). After adjusting political alignment for demographic effects, lifestyle politics decreased by 27.36% toward the political "center" and demographically confounded interests were no longer among the most polarized interests. Instead, after demographic deconfounding, we found that the most liberal interests included electric cars, Planned Parenthood, and liberal satire while the most conservative interests included the Republican Party and conservative commentators. We validate our measures of political alignment and lifestyle politics using the General Social Survey and find similar demographic entanglements with lifestyle politics existed before social media such as Facebook were ubiquitous, giving us strong confidence that our results are not due to echo chambers or filter bubbles. Likewise, since demographic characteristics exist prior to ideological values, we argue that the demographic confounding we observe is causally responsible for the extreme instances of lifestyle politics that we find among the aggregated interests. We conclude our paper by relating our results to Simpson's paradox, cultural omnivorousness, and network autocorrelation.

Hansheng Xue, Vijini Mallawaarachchi, Yujia Zhang et al.

Mixed communities of organisms are found in many environments (from the human gut to marine ecosystems) and can have profound impact on human health and the environment. Metagenomics studies the genomic material of such communities through high-throughput sequencing that yields DNA subsequences for subsequent analysis. A fundamental problem in the standard workflow, called binning, is to discover clusters, of genomic subsequences, associated with the unknown constituent organisms. Inherent noise in the subsequences, various biological constraints that need to be imposed on them and the skewed cluster size distribution exacerbate the difficulty of this unsupervised learning problem. In this paper, we present a new formulation using a graph where the nodes are subsequences and edges represent homophily information. In addition, we model biological constraints providing heterophilous signal about nodes that cannot be clustered together. We solve the binning problem by developing new algorithms for (i) graph representation learning that preserves both homophily relations and heterophily constraints (ii) constraint-based graph clustering method that addresses the problems of skewed cluster size distribution. Extensive experiments, on real and synthetic datasets, demonstrate that our approach, called RepBin, outperforms a wide variety of competing methods. Our constraint-based graph representation learning and clustering methods, that may be useful in other domains as well, advance the state-of-the-art in both metagenomics binning and graph representation learning.

Daping Zhang, Xin Chen, Yujia Zhang et al.

Agriculture is the fundamental industry of the society, which is the basis of food supply and an important source of employment and GDP increase. However, the insufficient expert can not fulfill the demand of farmers. To address this problem, we design a chatbot to answer frequently asked questions in the Agriculture field. Template-based questions will be answered by AIML while LSA is used for other service-based questions. This chatbot will assist farmers by dealing with industry problems conveniently and efficiently.

Yuzhi Zhao, Lai-Man Po, Wing-Yin Yu et al.

We propose a hybrid recurrent Video Colorization with Hybrid Generative Adversarial Network (VCGAN), an improved approach to video colorization using end-to-end learning. The VCGAN addresses two prevalent issues in the video colorization domain: Temporal consistency and unification of colorization network and refinement network into a single architecture. To enhance colorization quality and spatiotemporal consistency, the mainstream of generator in VCGAN is assisted by two additional networks, i.e., global feature extractor and placeholder feature extractor, respectively. The global feature extractor encodes the global semantics of grayscale input to enhance colorization quality, whereas the placeholder feature extractor acts as a feedback connection to encode the semantics of the previous colorized frame in order to maintain spatiotemporal consistency. If changing the input for placeholder feature extractor as grayscale input, the hybrid VCGAN also has the potential to perform image colorization. To improve the consistency of far frames, we propose a dense long-term loss that smooths the temporal disparity of every two remote frames. Trained with colorization and temporal losses jointly, VCGAN strikes a good balance between color vividness and video continuity. Experimental results demonstrate that VCGAN produces higher-quality and temporally more consistent colorful videos than existing approaches.

Yujia Zhang, Qianzhong Li, Xiaoguang Zhao et al.

Regular pavement inspection plays a significant role in road maintenance for safety assurance. Existing methods mainly address the tasks of crack detection and segmentation that are only tailored for long-thin crack disease. However, there are many other types of diseases with a wider variety of sizes and patterns that are also essential to segment in practice, bringing more challenges towards fine-grained pavement inspection. In this paper, our goal is not only to automatically segment cracks, but also to segment other complex pavement diseases as well as typical landmarks (markings, runway lights, etc.) and commonly seen water/oil stains in a single model. To this end, we propose a three-stream boundary-aware network (TB-Net). It consists of three streams fusing the low-level spatial and the high-level contextual representations as well as the detailed boundary information. Specifically, the spatial stream captures rich spatial features. The context stream, where an attention mechanism is utilized, models the contextual relationships over local features. The boundary stream learns detailed boundaries using a global-gated convolution to further refine the segmentation outputs. The network is trained using a dual-task loss in an end-to-end manner, and experiments on a newly collected fine-grained pavement disease dataset show the effectiveness of our TB-Net.

Jinting Wu, Yujia Zhang, Xiaoguang Zhao et al.

Although automatic emotion recognition from facial expressions and speech has made remarkable progress, emotion recognition from body gestures has not been thoroughly explored. People often use a variety of body language to express emotions, and it is difficult to enumerate all emotional body gestures and collect enough samples for each category. Therefore, recognizing new emotional body gestures is critical for better understanding human emotions. However, the existing methods fail to accurately determine which emotional state a new body gesture belongs to. In order to solve this problem, we introduce a Generalized Zero-Shot Learning (GZSL) framework, which consists of three branches to infer the emotional state of the new body gestures with only their semantic descriptions. The first branch is a Prototype-Based Detector (PBD) which is used to determine whether an sample belongs to a seen body gesture category and obtain the prediction results of the samples from the seen categories. The second branch is a Stacked AutoEncoder (StAE) with manifold regularization, which utilizes semantic representations to predict samples from unseen categories. Note that both of the above branches are for body gesture recognition. We further add an emotion classifier with a softmax layer as the third branch in order to better learn the feature representations for this emotion classification task. The input features for these three branches are learned by a shared feature extraction network, i.e., a Bidirectional Long Short-Term Memory Networks (BLSTM) with a self-attention module. We treat these three branches as subtasks and use multi-task learning strategies for joint training. The performance of our framework on an emotion recognition dataset is significantly superior to the traditional method of emotion classification and state-of-the-art zero-shot learning methods.

Jinting Wu, Yujia Zhang, Xiaoguang Zhao

Hand gesture recognition plays a significant role in human-computer interaction for understanding various human gestures and their intent. However, most prior works can only recognize gestures of limited labeled classes and fail to adapt to new categories. The task of Generalized Zero-Shot Learning (GZSL) for hand gesture recognition aims to address the above issue by leveraging semantic representations and detecting both seen and unseen class samples. In this paper, we propose an end-to-end prototype-based GZSL framework for hand gesture recognition which consists of two branches. The first branch is a prototype-based detector that learns gesture representations and determines whether an input sample belongs to a seen or unseen category. The second branch is a zero-shot label predictor which takes the features of unseen classes as input and outputs predictions through a learned mapping mechanism between the feature and the semantic space. We further establish a hand gesture dataset that specifically targets this GZSL task, and comprehensive experiments on this dataset demonstrate the effectiveness of our proposed approach on recognizing both seen and unseen gestures.

Yujia Zhang, Kuangyan Song, Yiming Sun et al.

Methods for interpreting machine learning black-box models increase the outcomes' transparency and in turn generates insight into the reliability and fairness of the algorithms. However, the interpretations themselves could contain significant uncertainty that undermines the trust in the outcomes and raises concern about the model's reliability. Focusing on the method "Local Interpretable Model-agnostic Explanations" (LIME), we demonstrate the presence of two sources of uncertainty, namely the randomness in its sampling procedure and the variation of interpretation quality across different input data points. Such uncertainty is present even in models with high training and test accuracy. We apply LIME to synthetic data and two public data sets, text classification in 20 Newsgroup and recidivism risk-scoring in COMPAS, to support our argument.

Yujia Zhang, Michael Kampffmeyer, Xiaodan Liang et al.

Video summarization plays an important role in video understanding by selecting key frames/shots. Traditionally, it aims to find the most representative and diverse contents in a video as short summaries. Recently, a more generalized task, query-conditioned video summarization, has been introduced, which takes user queries into consideration to learn more user-oriented summaries. In this paper, we propose a query-conditioned three-player generative adversarial network to tackle this challenge. The generator learns the joint representation of the user query and the video content, and the discriminator takes three pairs of query-conditioned summaries as the input to discriminate the real summary from a generated and a random one. A three-player loss is introduced for joint training of the generator and the discriminator, which forces the generator to learn better summary results, and avoids the generation of random trivial summaries. Experiments on a recently proposed query-conditioned video summarization benchmark dataset show the efficiency and efficacy of our proposed method.

Michael Kampffmeyer, Yinbo Chen, Xiaodan Liang et al.

Graph convolutional neural networks have recently shown great potential for the task of zero-shot learning. These models are highly sample efficient as related concepts in the graph structure share statistical strength allowing generalization to new classes when faced with a lack of data. However, multi-layer architectures, which are required to propagate knowledge to distant nodes in the graph, dilute the knowledge by performing extensive Laplacian smoothing at each layer and thereby consequently decrease performance. In order to still enjoy the benefit brought by the graph structure while preventing dilution of knowledge from distant nodes, we propose a Dense Graph Propagation (DGP) module with carefully designed direct links among distant nodes. DGP allows us to exploit the hierarchical graph structure of the knowledge graph through additional connections. These connections are added based on a node's relationship to its ancestors and descendants. A weighting scheme is further used to weigh their contribution depending on the distance to the node to improve information propagation in the graph. Combined with finetuning of the representations in a two-stage training approach our method outperforms state-of-the-art zero-shot learning approaches.

Yujia Zhang, Michael Kampffmeyer, Xiaodan Liang et al.

The large amount of videos popping up every day, make it more and more critical that key information within videos can be extracted and understood in a very short time. Video summarization, the task of finding the smallest subset of frames, which still conveys the whole story of a given video, is thus of great significance to improve efficiency of video understanding. We propose a novel Dilated Temporal Relational Generative Adversarial Network (DTR-GAN) to achieve frame-level video summarization. Given a video, it selects the set of key frames, which contain the most meaningful and compact information. Specifically, DTR-GAN learns a dilated temporal relational generator and a discriminator with three-player loss in an adversarial manner. A new dilated temporal relation (DTR) unit is introduced to enhance temporal representation capturing. The generator uses this unit to effectively exploit global multi-scale temporal context to select key frames and to complement the commonly used Bi-LSTM. To ensure that summaries capture enough key video representation from a global perspective rather than a trivial randomly shorten sequence, we present a discriminator that learns to enforce both the information completeness and compactness of summaries via a three-player loss. The loss includes the generated summary loss, the random summary loss, and the real summary (ground-truth) loss, which play important roles for better regularizing the learned model to obtain useful summaries. Comprehensive experiments on three public datasets show the effectiveness of the proposed approach.

Michael Kampffmeyer, Nanqing Dong, Xiaodan Liang et al.

Salient segmentation aims to segment out attention-grabbing regions, a critical yet challenging task and the foundation of many high-level computer vision applications. It requires semantic-aware grouping of pixels into salient regions and benefits from the utilization of global multi-scale contexts to achieve good local reasoning. Previous works often address it as two-class segmentation problems utilizing complicated multi-step procedures including refinement networks and complex graphical models. We argue that semantic salient segmentation can instead be effectively resolved by reformulating it as a simple yet intuitive pixel-pair based connectivity prediction task. Following the intuition that salient objects can be naturally grouped via semantic-aware connectivity between neighboring pixels, we propose a pure Connectivity Net (ConnNet). ConnNet predicts connectivity probabilities of each pixel with its neighboring pixels by leveraging multi-level cascade contexts embedded in the image and long-range pixel relations. We investigate our approach on two tasks, namely salient object segmentation and salient instance-level segmentation, and illustrate that consistent improvements can be obtained by modeling these tasks as connectivity instead of binary segmentation tasks for a variety of network architectures. We achieve state-of-the-art performance, outperforming or being comparable to existing approaches while reducing inference time due to our less complex approach.

Yujia Zhang, Xiaodan Liang, Dingwen Zhang et al.

Unsupervised video summarization plays an important role on digesting, browsing, and searching the ever-growing videos every day, and the underlying fine-grained semantic and motion information (i.e., objects of interest and their key motions) in online videos has been barely touched. In this paper, we investigate a pioneer research direction towards the fine-grained unsupervised object-level video summarization. It can be distinguished from existing pipelines in two aspects: extracting key motions of participated objects, and learning to summarize in an unsupervised and online manner. To achieve this goal, we propose a novel online motion Auto-Encoder (online motion-AE) framework that functions on the super-segmented object motion clips. Comprehensive experiments on a newly-collected surveillance dataset and public datasets have demonstrated the effectiveness of our proposed method.