Yujun Cai

Haoxuan Qu, Xiaofei Hui, Yujun Cai et al.

Open-set object recognition aims to identify if an object is from a class that has been encountered during training or not. To perform open-set object recognition accurately, a key challenge is how to reduce the reliance on spurious-discriminative features. In this paper, motivated by that different large models pre-trained through different paradigms can possess very rich while distinct implicit knowledge, we propose a novel framework named Large Model Collaboration (LMC) to tackle the above challenge via collaborating different off-the-shelf large models in a training-free manner. Moreover, we also incorporate the proposed framework with several novel designs to effectively extract implicit knowledge from large models. Extensive experiments demonstrate the efficacy of our proposed framework. Code is available https://github.com/Harryqu123/LMC

Yiwei Wang, Muhao Chen, Wenxuan Zhou et al.

Recent literature focuses on utilizing the entity information in the sentence-level relation extraction (RE), but this risks leaking superficial and spurious clues of relations. As a result, RE still suffers from unintended entity bias, i.e., the spurious correlation between entity mentions (names) and relations. Entity bias can mislead the RE models to extract the relations that do not exist in the text. To combat this issue, some previous work masks the entity mentions to prevent the RE models from overfitting entity mentions. However, this strategy degrades the RE performance because it loses the semantic information of entities. In this paper, we propose the CORE (Counterfactual Analysis based Relation Extraction) debiasing method that guides the RE models to focus on the main effects of textual context without losing the entity information. We first construct a causal graph for RE, which models the dependencies between variables in RE models. Then, we propose to conduct counterfactual analysis on our causal graph to distill and mitigate the entity bias, that captures the causal effects of specific entity mentions in each instance. Note that our CORE method is model-agnostic to debias existing RE systems during inference without changing their training processes. Extensive experimental results demonstrate that our CORE yields significant gains on both effectiveness and generalization for RE. The source code is provided at: https://github.com/vanoracai/CoRE.

Yiwei Wang, Yujun Cai, Muhao Chen et al.

Instruction-tuned large language models (LLMs), such as ChatGPT, have led to promising zero-shot performance in discriminative natural language understanding (NLU) tasks. This involves querying the LLM using a prompt containing the question, and the candidate labels to choose from. The question-answering capabilities of ChatGPT arise from its pre-training on large amounts of human-written text, as well as its subsequent fine-tuning on human preferences, which motivates us to ask: Does ChatGPT also inherits humans' cognitive biases? In this paper, we study the primacy effect of ChatGPT: the tendency of selecting the labels at earlier positions as the answer. We have two main findings: i) ChatGPT's decision is sensitive to the order of labels in the prompt; ii) ChatGPT has a clearly higher chance to select the labels at earlier positions as the answer. We hope that our experiments and analyses provide additional insights into building more reliable ChatGPT-based solutions. We release the source code at https://github.com/wangywUST/PrimacyEffectGPT.

Haoxuan Qu, Li Xu, Yujun Cai et al.

For tackling the task of 2D human pose estimation, the great majority of the recent methods regard this task as a heatmap estimation problem, and optimize the heatmap prediction using the Gaussian-smoothed heatmap as the optimization objective and using the pixel-wise loss (e.g. MSE) as the loss function. In this paper, we show that optimizing the heatmap prediction in such a way, the model performance of body joint localization, which is the intrinsic objective of this task, may not be consistently improved during the optimization process of the heatmap prediction. To address this problem, from a novel perspective, we propose to formulate the optimization of the heatmap prediction as a distribution matching problem between the predicted heatmap and the dot annotation of the body joint directly. By doing so, our proposed method does not need to construct the Gaussian-smoothed heatmap and can achieve a more consistent model performance improvement during the optimization of the heatmap prediction. We show the effectiveness of our proposed method through extensive experiments on the COCO dataset and the MPII dataset.

Shangchen Han, Po-chen Wu, Yubo Zhang et al.

Real-time tracking of 3D hand pose in world space is a challenging problem and plays an important role in VR interaction. Existing work in this space are limited to either producing root-relative (versus world space) 3D pose or rely on multiple stages such as generating heatmaps and kinematic optimization to obtain 3D pose. Moreover, the typical VR scenario, which involves multi-view tracking from wide \ac{fov} cameras is seldom addressed by these methods. In this paper, we present a unified end-to-end differentiable framework for multi-view, multi-frame hand tracking that directly predicts 3D hand pose in world space. We demonstrate the benefits of end-to-end differentiabilty by extending our framework with downstream tasks such as jitter reduction and pinch prediction. To demonstrate the efficacy of our model, we further present a new large-scale egocentric hand pose dataset that consists of both real and synthetic data. Experiments show that our system trained on this dataset handles various challenging interactive motions, and has been successfully applied to real-time VR applications.

Yiwei Wang, Muhao Chen, Wenxuan Zhou et al.

Entity types and textual context are essential properties for sentence-level relation extraction (RE). Existing work only encodes these properties within individual instances, which limits the performance of RE given the insufficient features in a single sentence. In contrast, we model these properties from the whole dataset and use the dataset-level information to enrich the semantics of every instance. We propose the GRAPHCACHE (Graph Neural Network as Caching) module, that propagates the features across sentences to learn better representations for RE. GRAPHCACHE aggregates the features from sentences in the whole dataset to learn global representations of properties, and use them to augment the local features within individual sentences. The global property features act as dataset-level prior knowledge for RE, and a complement to the sentence-level features. Inspired by the classical caching technique in computer systems, we develop GRAPHCACHE to update the property representations in an online manner. Overall, GRAPHCACHE yields significant effectiveness gains on RE and enables efficient message passing across all sentences in the dataset.

Yihong Luo, Siya Qiu, Xingjian Tao et al.

In this paper, we propose a novel generative model that utilizes a conditional Energy-Based Model (EBM) for enhancing Variational Autoencoder (VAE), termed Energy-Calibrated VAE (EC-VAE). Specifically, VAEs often suffer from blurry generated samples due to the lack of a tailored training on the samples generated in the generative direction. On the other hand, EBMs can generate high-quality samples but require expensive Markov Chain Monte Carlo (MCMC) sampling. To address these issues, we introduce a conditional EBM for calibrating the generative direction of VAE during training, without requiring it for the generation at test time. In particular, we train EC-VAE upon both the input data and the calibrated samples with adaptive weight to enhance efficacy while avoiding MCMC sampling at test time. Furthermore, we extend the calibration idea of EC-VAE to variational learning and normalizing flows, and apply EC-VAE to an additional application of zero-shot image restoration via neural transport prior and range-null theory. We evaluate the proposed method with two applications, including image generation and zero-shot image restoration, and the experimental results show that our method achieves competitive performance over single-step non-adversarial generation. Our code is available at https://github.com/DJ-LYH/EC-VAE.

Yiwei Wang, Bryan Hooi, Fei Wang et al.

Relation extraction (RE) aims to extract the relations between entity names from the textual context. In principle, textual context determines the ground-truth relation and the RE models should be able to correctly identify the relations reflected by the textual context. However, existing work has found that the RE models memorize the entity name patterns to make RE predictions while ignoring the textual context. This motivates us to raise the question: ``are RE models robust to the entity replacements?'' In this work, we operate the random and type-constrained entity replacements over the RE instances in TACRED and evaluate the state-of-the-art RE models under the entity replacements. We observe the 30\% - 50\% F1 score drops on the state-of-the-art RE models under entity replacements. These results suggest that we need more efforts to develop effective RE models robust to entity replacements. We release the source code at https://github.com/wangywUST/RobustRE.

Tao Wang, Jianfeng Zhang, Yujun Cai et al.

We present Multi-view Pose transformer (MvP) for estimating multi-person 3D poses from multi-view images. Instead of estimating 3D joint locations from costly volumetric representation or reconstructing the per-person 3D pose from multiple detected 2D poses as in previous methods, MvP directly regresses the multi-person 3D poses in a clean and efficient way, without relying on intermediate tasks. Specifically, MvP represents skeleton joints as learnable query embeddings and let them progressively attend to and reason over the multi-view information from the input images to directly regress the actual 3D joint locations. To improve the accuracy of such a simple pipeline, MvP presents a hierarchical scheme to concisely represent query embeddings of multi-person skeleton joints and introduces an input-dependent query adaptation approach. Further, MvP designs a novel geometrically guided attention mechanism, called projective attention, to more precisely fuse the cross-view information for each joint. MvP also introduces a RayConv operation to integrate the view-dependent camera geometry into the feature representations for augmenting the projective attention. We show experimentally that our MvP model outperforms the state-of-the-art methods on several benchmarks while being much more efficient. Notably, it achieves 92.3% AP25 on the challenging Panoptic dataset, improving upon the previous best approach [36] by 9.8%. MvP is general and also extendable to recovering human mesh represented by the SMPL model, thus useful for modeling multi-person body shapes. Code and models are available at https://github.com/sail-sg/mvp.

Haoxuan Qu, Yujun Cai, Jun Liu

Skeleton-based action recognition has attracted lots of research attention. Recently, to build an accurate skeleton-based action recognizer, a variety of works have been proposed. Among them, some works use large model architectures as backbones of their recognizers to boost the skeleton data representation capability, while some other works pre-train their recognizers on external data to enrich the knowledge. In this work, we observe that large language models which have been extensively used in various natural language processing tasks generally hold both large model architectures and rich implicit knowledge. Motivated by this, we propose a novel LLM-AR framework, in which we investigate treating the Large Language Model as an Action Recognizer. In our framework, we propose a linguistic projection process to project each input action signal (i.e., each skeleton sequence) into its ``sentence format'' (i.e., an ``action sentence''). Moreover, we also incorporate our framework with several designs to further facilitate this linguistic projection process. Extensive experiments demonstrate the efficacy of our proposed framework.

Sasha Salter, Richard Warren, Collin Schlager et al.

Hands are the primary means through which humans interact with the world. Reliable and always-available hand pose inference could yield new and intuitive control schemes for human-computer interactions, particularly in virtual and augmented reality. Computer vision is effective but requires one or multiple cameras and can struggle with occlusions, limited field of view, and poor lighting. Wearable wrist-based surface electromyography (sEMG) presents a promising alternative as an always-available modality sensing muscle activities that drive hand motion. However, sEMG signals are strongly dependent on user anatomy and sensor placement, and existing sEMG models have required hundreds of users and device placements to effectively generalize. To facilitate progress on sEMG pose inference, we introduce the emg2pose benchmark, the largest publicly available dataset of high-quality hand pose labels and wrist sEMG recordings. emg2pose contains 2kHz, 16 channel sEMG and pose labels from a 26-camera motion capture rig for 193 users, 370 hours, and 29 stages with diverse gestures - a scale comparable to vision-based hand pose datasets. We provide competitive baselines and challenging tasks evaluating real-world generalization scenarios: held-out users, sensor placements, and stages. emg2pose provides the machine learning community a platform for exploring complex generalization problems, holding potential to significantly enhance the development of sEMG-based human-computer interactions.

Hang Wu, Hongkai Chen, Yujun Cai et al.

Grounding natural language queries in graphical user interfaces (GUIs) poses unique challenges due to the diversity of visual elements, spatial clutter, and the ambiguity of language. In this paper, we introduce DiMo-GUI, a training-free framework for GUI grounding that leverages two core strategies: dynamic visual grounding and modality-aware optimization. Instead of treating the GUI as a monolithic image, our method splits the input into textual elements and iconic elements, allowing the model to reason over each modality independently using general-purpose vision-language models. When predictions are ambiguous or incorrect, DiMo-GUI dynamically focuses attention by generating candidate focal regions centered on the model's initial predictions and incrementally zooms into subregions to refine the grounding result. This hierarchical refinement process helps disambiguate visually crowded layouts without the need for additional training or annotations. We evaluate our approach on standard GUI grounding benchmarks and demonstrate consistent improvements over baseline inference pipelines, highlighting the effectiveness of combining modality separation with region-focused reasoning.

Zhecheng Li, Guoxian Song, Yujun Cai et al.

Modern Vision-Language Models (VLMs) exhibit remarkable visual and linguistic capabilities, achieving impressive performance in various tasks such as image recognition and object localization. However, their effectiveness in fine-grained tasks remains an open question. In everyday scenarios, individuals encountering design materials, such as magazines, typography tutorials, research papers, or branding content, may wish to identify aesthetically pleasing fonts used in the text. Given their multimodal capabilities and free accessibility, many VLMs are often considered potential tools for font recognition. This raises a fundamental question: Do VLMs truly possess the capability to recognize fonts? To investigate this, we introduce the Font Recognition Benchmark (FRB), a compact and well-structured dataset comprising 15 commonly used fonts. FRB includes two versions: (i) an easy version, where 10 sentences are rendered in different fonts, and (ii) a hard version, where each text sample consists of the names of the 15 fonts themselves, introducing a stroop effect that challenges model perception. Through extensive evaluation of various VLMs on font recognition tasks, we arrive at the following key findings: (i) Current VLMs exhibit limited font recognition capabilities, with many state-of-the-art models failing to achieve satisfactory performance and being easily affected by the stroop effect introduced by textual information. (ii) Few-shot learning and Chain-of-Thought (CoT) prompting provide minimal benefits in improving font recognition accuracy across different VLMs. (iii) Attention analysis sheds light on the inherent limitations of VLMs in capturing semantic features.

Zehao Li, Wenwei Han, Yujun Cai et al.

While 3D Gaussian Splatting enables high-quality real-time rendering, existing Gaussian-based frameworks for 3D semantic segmentation still face significant challenges in boundary recognition accuracy. To address this, we propose a novel 3DGS-based framework named GradiSeg, incorporating Identity Encoding to construct a deeper semantic understanding of scenes. Our approach introduces two key modules: Identity Gradient Guided Densification (IGD) and Local Adaptive K-Nearest Neighbors (LA-KNN). The IGD module supervises gradients of Identity Encoding to refine Gaussian distributions along object boundaries, aligning them closely with boundary contours. Meanwhile, the LA-KNN module employs position gradients to adaptively establish locality-aware propagation of Identity Encodings, preventing irregular Gaussian spreads near boundaries. We validate the effectiveness of our method through comprehensive experiments. Results show that GradiSeg effectively addresses boundary-related issues, significantly improving segmentation accuracy without compromising scene reconstruction quality. Furthermore, our method's robust segmentation capability and decoupled Identity Encoding representation make it highly suitable for various downstream scene editing tasks, including 3D object removal, swapping and so on.

Jianing Chen, Zehao Li, Yujun Cai et al.

Reconstructing dynamic 3D scenes from monocular videos remains a fundamental challenge in 3D vision. While 3D Gaussian Splatting (3DGS) achieves real-time rendering in static settings, extending it to dynamic scenes is challenging due to the difficulty of learning structured and temporally consistent motion representations. This challenge often manifests as three limitations in existing methods: redundant Gaussian updates, insufficient motion supervision, and weak modeling of complex non-rigid deformations. These issues collectively hinder coherent and efficient dynamic reconstruction. To address these limitations, we propose HAIF-GS, a unified framework that enables structured and consistent dynamic modeling through sparse anchor-driven deformation. It first identifies motion-relevant regions via an Anchor Filter to suppress redundant updates in static areas. A self-supervised Induced Flow-Guided Deformation module induces anchor motion using multi-frame feature aggregation, eliminating the need for explicit flow labels. To further handle fine-grained deformations, a Hierarchical Anchor Propagation mechanism increases anchor resolution based on motion complexity and propagates multi-level transformations. Extensive experiments on synthetic and real-world benchmarks validate that HAIF-GS significantly outperforms prior dynamic 3DGS methods in rendering quality, temporal coherence, and reconstruction efficiency.

Zhen Xiong, Yujun Cai, Zhecheng Li et al.

Recent advances in test-time scaling have enabled Large Language Models (LLMs) to display sophisticated reasoning abilities via extended Chain-of-Thought (CoT) generation. Despite their potential, these Reasoning LLMs (RLMs) often demonstrate counterintuitive and unstable behaviors, such as performance degradation under few-shot prompting, that challenge our current understanding of RLMs. In this work, we introduce a unified graph-based analytical framework for better modeling the reasoning processes of RLMs. Our method first clusters long, verbose CoT outputs into semantically coherent reasoning steps, then constructs directed reasoning graphs to capture contextual and logical dependencies among these steps. Through comprehensive analysis across models and prompting regimes, we reveal that structural properties, such as exploration density, branching, and convergence ratios, strongly correlate with reasoning accuracy. Our findings demonstrate how prompting strategies substantially reshape the internal reasoning structure of RLMs, directly affecting task outcomes. The proposed framework not only enables quantitative evaluation of reasoning quality beyond conventional metrics but also provides practical insights for prompt engineering and the cognitive analysis of LLMs. Code and resources will be released to facilitate future research in this direction.

Xingjian Tao, Yiwei Wang, Yujun Cai et al.

Multimodal large language models (MLLMs) have enabled GUI agents to interact with operating systems by grounding language into spatial actions. Despite their promising performance, these models frequently exhibit hallucinations-systematic localization errors that compromise reliability. We propose a fine-grained evaluation framework that categorizes model predictions into four distinct types, revealing nuanced failure modes beyond traditional accuracy metrics. To better quantify model uncertainty, we introduce the Peak Sharpness Score (PSS), a metric that evaluates the alignment between semantic continuity and logits distribution in coordinate prediction. Building on this insight, we further propose Context-Aware Cropping, a training-free technique that improves model performance by adaptively refining input context. Extensive experiments demonstrate that our framework and methods provide actionable insights and enhance the interpretability and robustness of GUI agent behavior.

Zehao Li, Hao Jiang, Yujun Cai et al.

Although dynamic scene reconstruction has long been a fundamental challenge in 3D vision, the recent emergence of 3D Gaussian Splatting (3DGS) offers a promising direction by enabling high-quality, real-time rendering through explicit Gaussian primitives. However, existing 3DGS-based methods for dynamic reconstruction often suffer from \textit{spatio-temporal incoherence} during initialization, where canonical Gaussians are constructed by aggregating observations from multiple frames without temporal distinction. This results in spatio-temporally entangled representations, making it difficult to model dynamic motion accurately. To overcome this limitation, we propose \textbf{STDR} (Spatio-Temporal Decoupling for Real-time rendering), a plug-and-play module that learns spatio-temporal probability distributions for each Gaussian. STDR introduces a spatio-temporal mask, a separated deformation field, and a consistency regularization to jointly disentangle spatial and temporal patterns. Extensive experiments demonstrate that incorporating our module into existing 3DGS-based dynamic scene reconstruction frameworks leads to notable improvements in both reconstruction quality and spatio-temporal consistency across synthetic and real-world benchmarks.

Liyang Chen, Yujun Cai, Jieqiong Dong et al.

Retrieval-Augmented Generation (RAG) systems require corpora that are both structurally clean and semantically coherent. BRIGHT is a recent and influential benchmark designed to evaluate complex multi-hop retrieval across diverse, high-reasoning domains. However, its practical effectiveness is limited by common web-crawled artifacts - such as content redundancy and semantic discontinuity - that impair retrieval accuracy and downstream reasoning. Notably, we find that such issues are concentrated in seven StackExchange-derived subdomains, while other domains (e.g., Coding and Theorem-based content) remain relatively clean. In this study, we present MARCUS, a multi-agent pipeline that leverages large language models (LLMs) to systematically clean and re-chunk BRIGHT into a higher-quality corpus: BRIGHT-Plus. MARCUS applies dedicated agents for structural noise removal and semantic segmentation, preserving answer-bearing spans while improving contextual integrity. Experimental evaluations demonstrate that BRIGHT-Plus yields consistent and significant improvements in both retrieval accuracy and multi-hop reasoning across a diverse set of retrievers. We release both the BRIGHT-Plus corpus and the MARCUS pipeline to support future research on robust, reasoning-centric retrieval.

Haoxuan Qu, Zhaoyang He, Zeyu Hu et al.

To facilitate the application of motion prediction in practice, recently, the few-shot motion prediction task has attracted increasing research attention. Yet, in existing few-shot motion prediction works, a specific model that is dedicatedly trained over human motions is generally required. In this work, rather than tackling this task through training a specific human motion prediction model, we instead propose a novel FMP-OC framework. In FMP-OC, in a totally training-free manner, we enable Few-shot Motion Prediction, which is a non-language task, to be performed directly via utilizing the Off-the-shelf language model ChatGPT. Specifically, to lead ChatGPT as a language model to become an accurate motion predictor, in FMP-OC, we first introduce several novel designs to facilitate extracting implicit knowledge from ChatGPT. Moreover, we also incorporate our framework with a motion-in-context learning mechanism. Extensive experiments demonstrate the efficacy of our proposed framework.

Zhe Sun, Yujun Cai, Jiayu Yao et al.

Large Audio-Language Models (LALMs) have recently shown impressive progress in speech recognition, audio captioning, and auditory question answering. Yet, whether these models can perceive spatial dynamics, particularly the motion of sound sources, remains unclear. In this work, we uncover a systematic motion perception deficit in current ALLMs. To investigate this issue, we introduce AMPBench, the first benchmark explicitly designed to evaluate auditory motion understanding. AMPBench introduces a controlled question-answering benchmark designed to evaluate whether Audio-Language Models (LALMs) can infer the direction and trajectory of moving sound sources from binaural audio. Comprehensive quantitative and qualitative analyses reveal that current models struggle to reliably recognize motion cues or distinguish directional patterns. The average accuracy remains below 50%, underscoring a fundamental limitation in auditory spatial reasoning. Our study highlights a fundamental gap between human and model auditory spatial reasoning, providing both a diagnostic tool and new insight for enhancing spatial cognition in future Audio-Language Models.

Fangyu Wu, Yujun Cai

Vision language models such as CLIP have shown remarkable performance in zero shot classification, but remain susceptible to spurious correlations, where irrelevant visual features influence predictions. Existing debiasing methods often require access to training data and explicit group labels to perform fine-tuning or adjust embeddings, which limits their practicality in real-world settings. Test-time methods attempt to avoid this constraint, but many still depend on prior knowledge of dataset specific biases, limiting their generalizability in open set settings. In this work, we propose a test-time debiasing method for ViT based CLIP models that requires no additional training or assumptions of bias annotations. Our approach uses a pretrained segmentation model to isolate the target visual attribute, then adjusts the non target regions so that their embeddings are uniformly similar to all class specific text prompts. This procedure removes unintended bias signals from confounding visual regions while preserving the target attribute. Experiments on Waterbirds and CelebA show that our method outperforms existing test-time debiasing approaches in both group robustness metrics and Attention IoU. These results demonstrate the effectiveness of segmentation guided interventions for scalable and annotation free bias mitigation in vision language models.

Yike Wu, Yiwei Wang, Yujun Cai

While Large Vision-Language Models (LVLMs) achieve strong performance in multimodal tasks, hallucinations continue to hinder their reliability. Among the three categories of hallucinations, which include object, attribute, and relation, relation hallucinations account for the largest proportion but have received the least attention. To address this issue, we propose ChainMPQ (Multi-Perspective Questions guided Interleaved Chain of Image and Text), a training-free method that improves relational inference in LVLMs by utilizing accumulated textual and visual memories. ChainMPQ first extracts subject and object keywords from the question to enhance the corresponding image regions. It then constructs multi-perspective questions that focus on the three core components of a relationship: the subject, the object, and the relation that links them. These questions are sequentially input to the model, with textual and visual memories from earlier steps providing supporting context for subsequent ones, thereby forming an interleaved chain of images and text that guides progressive relational reasoning. Experiments on multiple LVLMs and benchmarks show that ChainMPQ substantially reduces relation hallucinations, while ablation studies further validate the effectiveness of its three core modules.

Zhecheng Li, Guoxian Song, Yiwei Wang et al.

Grounding natural language queries in graphical user interfaces (GUIs) presents a challenging task that requires models to comprehend diverse UI elements across various applications and systems, while also accurately predicting the spatial coordinates for the intended operation. To tackle this problem, we propose GMS: Generalist Scanner Meets Specialist Locator, a synergistic coarse-to-fine framework that effectively improves GUI grounding performance. GMS leverages the complementary strengths of general vision-language models (VLMs) and small, task-specific GUI grounding models by assigning them distinct roles within the framework. Specifically, the general VLM acts as a 'Scanner' to identify potential regions of interest, while the fine-tuned grounding model serves as a 'Locator' that outputs precise coordinates within these regions. This design is inspired by how humans perform GUI grounding, where the eyes scan the interface and the brain focuses on interpretation and localization. Our whole framework consists of five stages and incorporates hierarchical search with cross-modal communication to achieve promising prediction results. Experimental results on the ScreenSpot-Pro dataset show that while the 'Scanner' and 'Locator' models achieve only $2.0\%$ and $3.7\%$ accuracy respectively when used independently, their integration within GMS framework yields an overall accuracy of $35.7\%$, representing a $10 \times$ improvement. Additionally, GMS significantly outperforms other strong baselines under various settings, demonstrating its robustness and potential for general-purpose GUI grounding.

Jingxuan Wei, Caijun Jia, Qi Chen et al.

Multimodal Machine Translation (MMT) enhances translation quality by incorporating visual context, helping to resolve textual ambiguities. While existing MMT methods perform well in bilingual settings, extending them to multilingual translation remains challenging due to cross-lingual interference and ineffective parameter-sharing strategies. To address this, we propose LLaVA-NeuMT, a novel multimodal multilingual translation framework that explicitly models language-specific and language-agnostic representations to mitigate multilingual interference. Our approach consists of a layer selection mechanism that identifies the most informative layers for different language pairs and a neuron-level adaptation strategy that dynamically selects language-specific and agnostic neurons to improve translation quality while reducing redundancy. We conduct extensive experiments on the M3-Multi30K and M3-AmbigCaps datasets, demonstrating that LLaVA-NeuMT, while fine-tuning only 40\% of the model parameters, surpasses full fine-tuning approaches and ultimately achieves SOTA results on both datasets. Our analysis further provides insights into the importance of selected layers and neurons in multimodal multilingual adaptation, offering an efficient and scalable solution to cross-lingual adaptation in multimodal translation.

Haoxuan Qu, Yujun Cai, Hossein Rahmani et al.

Recently, Gaussian Splatting (GS) has received a lot of attention in surface reconstruction. However, while 3D objects can be of complex and diverse shapes in the real world, existing GS-based methods only limitedly use a single type of splatting primitive (Gaussian ellipse or Gaussian ellipsoid) to represent object surfaces during their reconstruction. In this paper, we highlight that this can be insufficient for object surfaces to be represented in high quality. Thus, we propose a novel framework that, for the first time, enables Gaussian Splatting to incorporate multiple types of (geometrical) primitives during its surface reconstruction process. Specifically, in our framework, we first propose a compositional splatting strategy, enabling the splatting and rendering of different types of primitives in the Gaussian Splatting pipeline. In addition, we also design our framework with a mixed-primitive-based initialization strategy and a vertex pruning mechanism to further promote its surface representation learning process to be well executed leveraging different types of primitives. Extensive experiments show the efficacy of our framework and its accurate surface reconstruction performance.

Sifan Li, Yujun Cai, Yiwei Wang

Vision-language models (VLMs) excel in semantic tasks but falter at a core human capability: detecting hidden content in optical illusions or AI-generated images through perceptual adjustments like zooming. We introduce HC-Bench, a benchmark of 112 images with hidden text, objects, and illusions, revealing that leading VLMs achieve near-zero accuracy (0-5.36%)-even with explicit prompting. Humans resolve such ambiguities instinctively, yet VLMs fail due to an overreliance on high-level semantics. Strikingly, we propose SemVink (Semantic Visual Thinking) by simply scaling images to low resolutions (32-128 pixels), which unlocks >99% accuracy by eliminating redundant visual noise. This exposes a critical architectural flaw: VLMs prioritize abstract reasoning over low-level visual operations crucial for real-world robustness. Our work urges a shift toward hybrid models integrating multi-scale processing, bridging the gap between computational vision and human cognition for applications in medical imaging, security, and beyond.

Yiwei Wang, Yujun Cai, Yuxuan Liang et al.

We present a new neighbor sampling method on temporal graphs. In a temporal graph, predicting different nodes' time-varying properties can require the receptive neighborhood of various temporal scales. In this work, we propose the TNS (Time-aware Neighbor Sampling) method: TNS learns from temporal information to provide an adaptive receptive neighborhood for every node at any time. Learning how to sample neighbors is non-trivial, since the neighbor indices in time order are discrete and not differentiable. To address this challenge, we transform neighbor indices from discrete values to continuous ones by interpolating the neighbors' messages. TNS can be flexibly incorporated into popular temporal graph networks to improve their effectiveness without increasing their time complexity. TNS can be trained in an end-to-end manner. It needs no extra supervision and is automatically and implicitly guided to sample the neighbors that are most beneficial for prediction. Empirical results on multiple standard datasets show that TNS yields significant gains on edge prediction and node classification.

Mingfei Chen, Jianfeng Zhang, Xiangyu Xu et al.

In this work we develop a generalizable and efficient Neural Radiance Field (NeRF) pipeline for high-fidelity free-viewpoint human body synthesis under settings with sparse camera views. Though existing NeRF-based methods can synthesize rather realistic details for human body, they tend to produce poor results when the input has self-occlusion, especially for unseen humans under sparse views. Moreover, these methods often require a large number of sampling points for rendering, which leads to low efficiency and limits their real-world applicability. To address these challenges, we propose a Geometry-guided Progressive NeRF (GP-NeRF). In particular, to better tackle self-occlusion, we devise a geometry-guided multi-view feature integration approach that utilizes the estimated geometry prior to integrate the incomplete information from input views and construct a complete geometry volume for the target human body. Meanwhile, for achieving higher rendering efficiency, we introduce a progressive rendering pipeline through geometry guidance, which leverages the geometric feature volume and the predicted density values to progressively reduce the number of sampling points and speed up the rendering process. Experiments on the ZJU-MoCap and THUman datasets show that our method outperforms the state-of-the-arts significantly across multiple generalization settings, while the time cost is reduced > 70% via applying our efficient progressive rendering pipeline.

Yiwei Wang, Yujun Cai, Yuxuan Liang et al.

Representing a label distribution as a one-hot vector is a common practice in training node classification models. However, the one-hot representation may not adequately reflect the semantic characteristics of a node in different classes, as some nodes may be semantically close to their neighbors in other classes. It would cause over-confidence since the models are encouraged to assign full probabilities when classifying every node. While training models with label smoothing can ease this problem to some degree, it still fails to capture the nodes' semantic characteristics implied by the graph structures. In this work, we propose a novel SALS (\textit{Structure-Aware Label Smoothing}) method as an enhancement component to popular node classification models. SALS leverages the graph structures to capture the semantic correlations between the connected nodes and generate the structure-aware label distribution to replace the original one-hot label vectors, thus improving the node classification performance without inference costs. Extensive experiments on seven node classification benchmark datasets reveal the effectiveness of our SALS on improving both transductive and inductive node classification. Empirical results show that SALS is superior to the label smoothing method and enhances the node classification models to outperform the baseline methods.

Yiwei Wang, Wei Wang, Yuxuan Liang et al.

We present a new method to regularize graph neural networks (GNNs) for better generalization in graph classification. Observing that the omission of sub-structures does not necessarily change the class label of the whole graph, we develop the \textbf{GraphCrop} (Subgraph Cropping) data augmentation method to simulate the real-world noise of sub-structure omission. In principle, GraphCrop utilizes a node-centric strategy to crop a contiguous subgraph from the original graph while maintaining its connectivity. By preserving the valid structure contexts for graph classification, we encourage GNNs to understand the content of graph structures in a global sense, rather than rely on a few key nodes or edges, which may not always be present. GraphCrop is parameter learning free and easy to implement within existing GNN-based graph classifiers. Qualitatively, GraphCrop expands the existing training set by generating novel and informative augmented graphs, which retain the original graph labels in most cases. Quantitatively, GraphCrop yields significant and consistent gains on multiple standard datasets, and thus enhances the popular GNNs to outperform the baseline methods.

Chi Zhang, Yujun Cai, Guosheng Lin et al.

In this work, we develop methods for few-shot image classification from a new perspective of optimal matching between image regions. We employ the Earth Mover's Distance (EMD) as a metric to compute a structural distance between dense image representations to determine image relevance. The EMD generates the optimal matching flows between structural elements that have the minimum matching cost, which is used to calculate the image distance for classification. To generate the important weights of elements in the EMD formulation, we design a cross-reference mechanism, which can effectively alleviate the adverse impact caused by the cluttered background and large intra-class appearance variations. To implement k-shot classification, we propose to learn a structured fully connected layer that can directly classify dense image representations with the EMD. Based on the implicit function theorem, the EMD can be inserted as a layer into the network for end-to-end training. Our extensive experiments validate the effectiveness of our algorithm which outperforms state-of-the-art methods by a significant margin on five widely used few-shot classification benchmarks, namely, miniImageNet, tieredImageNet, Fewshot-CIFAR100 (FC100), Caltech-UCSD Birds-200-2011 (CUB), and CIFAR-FewShot (CIFAR-FS). We also demonstrate the effectiveness of our method on the image retrieval task in our experiments.