Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Rong ZhangSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works187 from public data
- Learning Affinity from Attention: End-to-End Weakly-Supervised Semantic Segmentation with Transformers287
An Affinity from Attention (AFA) module is proposed to learn semantic affinity from the multi-head self-attention (MHSA) in Transformers, which naturally integrate global information, to generate more integral initial pseudo labels for end-to-end WSSS.
- Token Contrast for Weakly-Supervised Semantic Segmentation213
Experiments on the PASCAL VOC and MS COCO datasets show the proposed ToCo can remarkably surpass other single-stage competitors and achieve comparable performance with state-of-the-art multi-stage methods.
- Deep Graph-neighbor Coherence Preserving Network for Unsupervised Cross-modal Hashing207
A deep graph-neighbor coherence preserving network (DGCPN), which addresses the inaccurate similarity problem by exploring and exploiting the data's intrinsic relationships in a graph and regulates comprehensive similarity preserving losses.
- Contrastive Boundary Learning for Point Cloud Segmentation204
This paper proposes a novel contrastive boundary learning (CBL) framework for point cloud segmentation that enhances feature discrimination between points across boundaries by contrasting their representations with the assistance of scene contexts at multiple scales.
- Expression snippet transformer for robust video-based facial expression recognition119
This work designs an attention-augmented snippet feature extractor (AA-SFE) to enhance the encoding of subtle facial movements of each snippet by gradually attending to more salient information, and introduces a shuffled snippet order prediction (SSOP) head and a corresponding loss to improve the modeling of subtle motion changes across subsequent snippets.
- Comprehensive Linguistic-Visual Composition Network for Image Retrieval98
A Comprehensive Linguistic-Visual Composition Network (CLVC-Net) for image retrieval is proposed, targeting comprehensive multi-modal compositions and a mutual enhancement module is designed to promote local-wise and global-wise composition processes by forcing them to share knowledge with each other.
- Estimating Noise Transition Matrix with Label Correlations for Noisy Multi-Label Learning95
This work studies the identifiability problem of the class-dependent transition matrix in noisy multi-label learning, and inspired by the identifiability results, a new estimator is proposed by exploiting label correlations without neither anchor points nor accurate class posterior, leading to better classification performance.
- On Exploring Undetermined Relationships for Visual Relationship Detection92
This paper proposes a novel multi-modal feature based undetermined relationship learning network (MF-URLN) and achieves great improvements in relationship detection, compared with state-of-the-art methods.
- Harnessing the Power of MLLMs for Transferable Text-to-Image Person ReID91
A novel method is introduced that automatically identifies words in a description that do not correspond with the image, based on the similarity between one text and all patch token embeddings in the image, which alleviates the impact of noisy textual descriptions.
- 85
- 84
- Parameter Efficient Multi-task Model Fusion with Partial Linearization79
This work proposes a novel method to improve multi-task fusion for parameter-efficient fine-tuning techniques like LoRA fine-tuning, which partially linearizes only the adapter modules and applies task arithmetic over the linearized adapters.
- 79
- 69
- Weakly-Supervised Semantic Segmentation with Visual Words Learning and Hybrid Pooling64
The second drawback of CAMs is alleviated with the proposed hybrid pooling, which incorporates the global average and local discriminative information to simultaneously ensure object completeness and reduce background regions.
- Multi-target Knowledge Distillation via Student Self-reflection63
A newknowledge distillation framework entitled multi-target knowledge distillation via student self-reflection or MTKD-SSR, which can not only enhance the teacher's ability in unfolding the knowledge to be distilled, but also improve the student’s capacity of digesting the knowledge.
- Gapformer: Graph Transformer with Graph Pooling for Node Classification61
Gapformer is presented, a method for node classification that deeply incorporates Graph Transformer with Graph Pooling and coarsens the large-scale nodes of a graph into a smaller number of pooling node via local or global graph pooling methods, and then computes the attention solely with the pooling nodes rather than all other nodes.
- Resistance Training Using Prior Bias: Toward Unbiased Scene Graph Generation61
Resistance Training using Prior Bias (RTPB) uses a distributed-based prior bias to improve models' detecting ability on less frequent relationships during training, thus improving the model generalizability on tail categories.
- Comprehensive Graph Gradual Pruning for Sparse Training in Graph Neural Networks59
A comprehensive graph gradual pruning framework termed CGP is proposed, designed by designing a during-training graph pruning paradigm to dynamically prune GNNs within one training process, which significantly reduces the computation costs.
- CLNode: Curriculum Learning for Node Classification55
Experimental results on real-world networks demonstrate that CLNode is a general framework that can be combined with various GNNs to improve their accuracy and robustness.
- HL-Net: Heterophily Learning Network for Scene Graph Generation54
A novel Heterophily Learning Network (HL-Net) is proposed to comprehensively explore the homophily and heterophily be-tween objects/relationships in scene graphs and demonstrates the superiority of the proposed HL-Net over existing state-of-the-art approaches.
- DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes49
The DIVOTrack is introduced: a novel baseline cross-view tracking method with a unified joint detection and cross-view tracking framework named CrossMOT, which learns object detection, single-view association, and cross-view matching with an all-in-one embedding model.
- RU-Net: Regularized Unrolling Network for Scene Graph Generation48
This paper first study the relation between GMP and graph Laplacian denoising (GLD) from the perspective of the unrolling technique, determining that GMP can be formulated as a solver for GLD, and proposes an unrolled message passing module and an fp-based graph regularization to suppress spurious connections between nodes.
- 48
- 47
- Toward Efficient Language Model Pretraining and Downstream Adaptation via Self-Evolution: A Case Study on SuperGLUE46
This technical report briefly describes the JDExplore d-team's Vega v2 submission on the SuperGLUE leaderboard, which achieved new state-of-the-art performance on 4/8 tasks, and the prompt transfer technique to improve the low-resource tasks by transferring the knowledge from the foundation model and related downstream tasks to the target task.
- 44
- Healthcare Copilot: Eliciting the Power of General LLMs for Medical Consultation40
Extensive results demonstrate that the proposed Healthcare Copilot significantly enhances the capabilities of general LLMs for medical consultations in terms of inquiry capability, conversational fluency, response accuracy, and safety.
- 39
- Where Does the Performance Improvement Come From?39
The reproducibility of image-text retrieval models is analyzed and the examination of the elements that lead to improved performance by pretrained and nonpretrained models in retrieving images and text are examined.
- Pose-disentangled Contrastive Learning for Self-supervised Facial Representation38
This work proposes a novel Pose-disentangled Contrastive Learning (PCL) method that learns pose-related information based on data augmentation of the same image, which would deliver more effective face-aware representation for various downstream tasks.
- Multi-task Compositional Network for Visual Relationship Detection37
This paper proposes a novel multi-task compositional network (MCN) that simultaneously performs object detection, predicate detection, and significance detection and proposes a multimodal feature fusion strategy based on visual, spatial, and label features and a novel correlated loss function.
- Sample-Cohesive Pose-Aware Contrastive Facial Representation Learning34
A Pose-disentangled Contrastive Facial Representation Learning (PCFRL) framework to enhance pose awareness for SFRL by explicitly disentangled the pose-aware features from non-pose face-aware features and introducing appropriate sample calibration schemes for better CL with the disentangled features.
- Noise-Resistant Multimodal Transformer for Emotion Recognition34
A novel paradigm that attempts to extract noise-resistant features in its pipeline and introduces a noise-aware learning scheme to effectively improve the robustness of multimodal emotion understanding against noisy information is presented.
- 33
- Confronting Reward Overoptimization for Diffusion Models: A Perspective of Inductive and Primacy Biases32
This work confronts the reward overoptimization problem in diffusion model alignment through the lenses of both inductive and primacy biases, and proposes Temporal Diffusion Policy Optimization with critic active neuron Reset (TDPO-R), a policy gradient algorithm that exploits the temporal inductive bias of diffusion models and mitigates the primacy bias stemming from active neurons.
- Comprehensive Distance-Preserving Autoencoders for Cross-Modal Retrieval32
Compared with eight state-of-the-art cross-modal retrieval methods, the experimental results demonstrate that the CDPAE outperforms all the unsupervised methods and performs competitively with the supervised methods.
- Gradformer: Graph Transformer with Exponential Decay30
Gradformer is presented, a method innovatively integrating GT with the intrinsic inductive bias by applying an exponential decay mask to the attention matrix, which enables Gradformer to retain its ability to capture information from distant nodes while focusing on the graph's local details.
- Image quality assessment using a SVD-based structural projection30
The accuracy, consistency, robustness, and stability of the proposed IQA model compared to state-of-the-art IQA methods, such as Visual Information Fidelity (VIF), Visual Signal to Noise Ratio (VSNR), and Structural Similarity Index (SSIM), are demonstrated.
- SkipNode: On Alleviating Performance Degradation for Deep Graph Convolutional Networks29
This work proposes a simple yet effective plug-and-play module, SkipNode, to overcome the performance degradation of deep GCNs and provides both theoretical analysis and empirical evaluation to demonstrate the efficacy of SkipNode and its superiority over SOTA baselines.
- Improving Adversarial Robustness via Mutual Information Estimation26
This paper investigates the dependence between outputs of the target model and input adversarial samples from the perspective of information theory, and proposes an adversarial defense method that could effectively improve the adversarial accuracy against multiple attacks.
- 26
- Empowering Embodied Manipulation: A Bimanual-Mobile Robot Manipulation Dataset for Household Tasks25
A novel Manipulation Efficiency Score (MES) metric is introduced to evaluate both the precision and efficiency of robot manipulation methods in household tasks using the BRMData, aiming to drive the development of bimanual-mobile robot manipulation technologies.
- 25
- Collocation and Try-on Network24
A Collocation and Try-On Network (CTO-Net) for FCM is proposed, combining both the collocation and try-on compatibilities, and it is demonstrated that the proposed CTO-Net significantly outperforms the state-of-the-art methods.
- Attribute-Image Person Re-identification via Modal-Consistent Metric Learning21
This paper proposes a modal-consistent metric learning (MCML) method that stably measures comprehensive similarities between attributes and images and integrates the CMMT loss function and MCMR, requiring no complex cross-modal feature alignments.
- Healthcare agent: eliciting the power of large language models for medical consultation20
The results demonstrate that the healthcare agent significantly enhances the capabilities of general LLMs in medical consultations, particularly in inquiry quality, response quality, and safety.
- Multi-Step Denoising Scheduled Sampling: Towards Alleviating Exposure Bias for Diffusion Models20
A multi-step denoising scheduled sampling (MDSS) strategy to alleviate the exposure bias in DDPMs and demonstrates that this approach is more effective in mitigating exposure bias in DDPM, DDIM, and DPM-solver.
- 20
- Open-Set Video-based Facial Expression Recognition with Human Expression-sensitive Prompting18
This paper proposes a novel Human Expression-Sensitive Prompting (HESP) mechanism to significantly enhance CLIP's ability to model video-based facial expression details effectively, and demonstrates that HESP can significantly boost CLIP's performance and outperform other state-of-the-art open-set video understanding methods by a large margin.
- KE-X: Towards subgraph explanations of knowledge graph embedding based on knowledge information gain18
- Bag of Tricks for Effective Language Model Pretraining and Downstream Adaptation: A Case Study on GLUE18
The JDExplore d-team's Vega v1 is the first to exceed powerful human performance on the two challenging tasks, i.e., SST-2 and WNLI, and believes its empirically successful recipe with a bag of tricks could shed new light on developing efficient discriminative large language models.
- 18
- 18
- IoUformer: Pseudo-IoU prediction with transformer for visual tracking17
A novel simple tracking paradigm based on the intersection over union (IoU) value prediction, which can predict reliable IoU values near-linear with the ground truth, which paves a safe way for the new IoU-based siamese tracking paradigm.
- Vega-MT: The JD Explore Academy Translation System for WMT2217
The JD Explore Academy’s submission of the WMT 2022 shared general translation task is described, which pushes the limit of bidirectional training for translation by scaling up two main factors, i.e. language pairs and model sizes.
- 17
- 16
- Chasing Consistency in Text-to-3D Generation from a Single Image16
Experimental results demonstrate that Consist3D produces more consistent, faithful, and photo-realistic 3D assets compared to previous state-of-the-art methods.
- BatchFormerV2: Exploring Sample Relationships for Dense Representation Learning16
A more general batch Transformer module, BatchFormerV2, which further enables exploring sample relationships for dense representation learning and consistently improves current DETR-based detection methods by over 1.3%.
- 16
- Semantics-Oriented Multitask Learning for DeepFake Detection: A Joint Embedding Approach15
This paper proposes an automated dataset expansion technique that broadens current face forgery datasets to support semantics-oriented DeepFake detection tasks at both the global face attribute and local face region levels, and resorts to the joint embedding of face images and labels for prediction.
- 15
- ChartMaster: Advancing Chart-to-Code Generation with Real-World Charts and Chart Similarity Reinforcement Learning15
The ChartMaster model, a GRPO-based reinforcement learning algorithm guided by a novel chart similarity reward, is developed, achieving SOTA results among 7B-parameter models and rivaling GPT-4o on various chart-to-code benchmarks.
- 15
- 15
- 15
- Semantic Contextualization of Face Forgery: A New Definition, Dataset, and Detection Method14
A semantics-oriented face forgery detection method that captures label relations and prioritizes the primary task (i.e., real or fake face detection) is proposed and shown to be superior over traditional binary and multi-class classification-based detectors.
- MuEP: A Multimodal Benchmark for Embodied Planning with Foundation Models14
Experimental results show that foundation models based on textual representations of environments usually outperform their visual counterparts, suggesting a gap in embodied planning abilities with multimodal observations.
- 14
- Semantics-Enriched Cross-Modal Alignment for Complex-Query Video Moment Retrieval13
This paper presents a novel solution termed semantics-enriched video moment retrieval method (SVMR), which can effectively and explicitly model the hierarchical multi-granularity semantics of complex textual query and explores cross-token relations to offer multiple granularity query representations with hierarchical semantic contexts of semantically associated tokens.
- 13
- Dual-branch Density Ratio Estimation for Signed Network Embedding13
Extensive empirical results demonstrate that DDRE not only significantly outperforms the methods based on density ratio estimation but also achieves competitive performance compared with other types of methods such as graph likelihood, generative adversarial networks, and graph convolutional networks.
- 13
- 13
- SpliceMix: A Cross-Scale and Semantic Blending Augmentation Strategy for Multi-Label Image Classification12
This article introduces a simple but effective augmentation strategy for multi-label image classification, namely SpliceMix, and provides a simple and non-parametric extension based on consistency learning (SpliceMix-CL) to show the potential of extending the authors' SpliceMix.
- 12
- 12
- 12
- DreamBooth++: Boosting Subject-Driven Generation via Region-Level References Packing12
This work introduces DreamBooth++, a region-level training strategy designed to significantly improve the efficiency and effectiveness of learning specific subjects, and employs a region-level data re-formulation technique that packs a set of reference images into a single sample, significantly reducing computational costs.
- TD²-Net: Toward Denoising and Debiasing for Video Scene Graph Generation12
This paper introduces a network named TD2-Net that aims at denoising and debiasing for dynamic SGG and proposes a denoising spatio-temporal transformer module that enhances object representation with robust contextual information.
- 12
- 11
- An Atomic Skill Library Construction Method for Data-Efficient Embodied Manipulation11
This work introduces a three-wheeled data-driven method to build an atomic skill library and shifts focus from end-to-end tasks to atomic skills, significantly reducing data costs while maintaining high performance and enabling efficient adaptation to new tasks.
- NoVo: Norm Voting off Hallucinations with Attention Heads in Large Language Models11
NoVo is presented, which harnesses the untapped potential of attention head norms to dramatically enhance factual accuracy in zero-shot multiple-choice questions (MCQs) and opens new frontiers in LLM interpretability, robustness and reliability.
- 11
- Region-adaptive Concept Aggregation for Few-shot Visual Recognition11
A novel metric-based FSL method termed region-adaptive concept aggregation network or RCA-Net is proposed, which devise a region- Adaptive concept aggregator (RCA) to model the relationships of different regions and capture the conceptual information in different regions, which are then integrated in a weighted average manner to obtain the conceptual representation.
- Not All Instances Contribute Equally: Instance-Adaptive Class Representation Learning for Few-Shot Visual Recognition11
This work develops an adaptive instance revaluing network (AIRN) with the capability to address the biased representation issue when generating the class representation, by learning and assigning adaptive weights for different instances according to their relative significance in the support set of corresponding class.
- End-to-End HOI Reconstruction Transformer with Graph-based Encoding10
This work proposes the End-to-End HOI Reconstruction Transformer with Graph-based Encoding (HOI-TG), which implicitly learns the interaction between humans and objects by leveraging self-attention mechanisms and achieves state-of-the-art performance on BEHAVE and InterCap datasets.
- On neural architecture search and hyperparameter optimization: A max-flow based approach10
A novel max-flow based search algorithm for AutoML is proposed by representing NAS and HPO as a Max-Flow problem on a graph and thus a couple of novel AutoML strategies, dubbed MF-NAS and MF-HPO, which handle the search space and the search strategy graphically are derived.
- 10
- Finding core labels for maximizing generalization of graph neural networks10
This paper introduces and elaborate on the problem of training data selection for GNNs, and proposes the following subset hypothesis for graph data: "There exists a core subset when selecting a fixed-size dataset from the dense training dataset, and GNNs trained on this core subset can achieve a better graph representation".
- 10
- 10
- 9
- Omniforce: on human-centered, large model empowered and cloud-edge collaborative AutoML system9
OmniForce is introduced, a human-centered AutoML system that yields both human-assisted ML and ML-assisted human techniques, putting an AutoML system into practice and building AI in open-environment scenarios, such as using large models in industrial supply chains.
- Graph structure reforming framework enhanced by commute time distance for graph classification9
This work proposes a plug-and-play framework based on Commute Time Distance (CTD), in which information is propagated in commute time distance neighborhoods and can capture higher-order structural information by utilizing commute paths to enhance the expressive power of GNNs.
- 8
- AGMixup: Adaptive Graph Mixup for Semi-supervised Node Classification8
An Adaptive Graph Mixup framework for semi-supervised node classification that introduces a subgraph-centric approach, which treats each subgraph similarly to how images are handled in Euclidean domains, thus facilitating a more natural integration of mixup into graph-based learning.
- Masked Graph Auto-Encoder Constrained Graph Pooling8
The experimental results reveal that MGAP has the capacity to consistently improve the performance of all the three node drop pooling models in the graph classification task and is a plug-and-play technique that can be directly adopted in the current node drop pooling methods.
- Bi-Level Optimization for Self-Supervised AI-Generated Face Detection7
A self-supervised method based on bi-level optimization that significantly outperform existing approaches in both one-class and binary classification settings, exhibiting strong generalization to unseen generators.
- Self-Supervised Learning for Detecting AI-Generated Faces as Anomalies7
An anomaly detection method for AI-generated faces is described by leveraging self-supervised learning of camera-intrinsic and face-specific features purely from photographic face images using a Gaussian mixture model.
- Building accurate translation-tailored large language models with language-aware instruction tuning7
A two-stage fine-tuning algorithm to improve the instruction-following ability of translation-tailored LLMs, particularly for maintaining accurate translation directions, and shows that it can preserve the model’s performance on other tasks, such as supervised translation and general tasks.
- 7
- 7
- 7
- 7
- DA-MoE: Addressing depth-sensitivity in graph-level analysis through mixture of experts6
The depth adaptive mixture of expert (DA-MoE) method is proposed, which incorporates two main improvements to GNN backbone: 1) DA-MoE employs different GNN layers, each considered an expert with its own parameters, effectively addressing the depth-sensitivity issue in graph data.
- Towards Alleviating Text-to-Image Retrieval Hallucination for CLIP in Zero-shot Learning6
This work proposes a Balanced Similarity with Auxiliary Prompts (BSAP) to mitigate the text-to-image retrieval bias of CLIP and attempts to adopt a hybrid similarity that combines the BSAP with the original similarity of CLIP to obtain a more robust outcome.
- Large Language Models as an Indirect Reasoner: Contrapositive and Contradiction for Automated Reasoning6
A Direct-Indirect Reasoning (DIR) method is proposed, which considers DR and IR as multiple parallel reasoning paths that are merged to derive the final answer and demonstrates that this method significantly outperforms all the original methods.
- 6
- Attentional Composition Networks for Long-Tailed Human Action Recognition6
A compositional learning based solution for video-based human action recognition that first learns verb-like and preposition-like components, then shuffles these components to generate samples for the tail classes in the feature space to augment the data for the Tail classes.
- Smile on the Face, Sadness in the Eyes: Bridging the Emotion Gap With a Multimodal Dataset of Eye and Facial Behaviors5
A comprehensive analysis of the importance of eye behaviors in ER is provided, advancing the study on addressing the gap between FER and ER for more robust ER performance.
- 5
- 5
- 5
- 5
- Multimodal Feature Analysis for Surgery Duration Predication5
The proposed multimodal feature fusion method outperforms the second competitor in terms of MAE, RMSE, MAPE, MAPE, and R2 by 4.6%, 5.9%, 6.0%, and 34.7%, respectively, demonstrating the proposed method’s superiority.
- Free-Form Composition Networks for Egocentric Action Recognition5
A free-form composition network (FFCN) that can simultaneously learn disentangled verb, preposition, and noun representations, and then use them to compose new samples in the feature space for rare classes of action videos is proposed.
- Unlikelihood Tuning on Negative Samples Amazingly Improves Zero-Shot Translation5
The authors' analyses suggest that although they work well in ideal ON settings, language IDs become fragile and lose their navigation ability when faced with off-target tokens, which commonly exist during inference but are rare in training scenarios.
- 5
- SAR target classification based on multiscale sparse representation5
This work firstly extracts the dense SIFT descriptors on multiple scales, then trains a global multiscale dictionary by sparse coding algorithm, which can provide more information and descriptive ability than single-scale ones.
- 5
- Reverse Prompt: Cracking the Recipe Inside Text-to-Image Generation4
This paper explores how to decode textual prompts from reference images, a process the authors refer to as image reverse prompt engineering, and proposes a method known as automatic reverse prompt optimization (ARPO), which refines an initial prompt into a high-quality prompt through an iteratively imitative gradient prompt optimization process.
- 4
- Towards Training A Chinese Large Language Model for Anesthesiology4
Hypnos, a Chinese Anesthesia model built upon existing LLMs, outperforms other medical LLMs in anesthesiology in metrics, GPT-4, and human evaluation on the benchmark dataset.
- HAG-Former: A Temporal-Polarimetric Relationship Inference Network From Local to Global4
A novel Hybrid Attention-GRU Transformer (HAG-Former) model is introduced, which harnesses the power of pixel-level temporal-polarimetric change analysis and captures the dynamic variations in polarization scattering properties, thereby enhancing classification robustness and versatility.
- Graph explicit pooling for graph-level representation learning4
A novel Graph explicit Pooling (GrePool) method is introduced, which selects nodes by explicitly leveraging the relationships between the nodes and final representation vectors crucial for classification, which outperforms 14 baseline methods for most datasets and enhances GrePool's performance without incurring additional computational costs.
- 4
- 3
- 3
- 3
- Smile upon the Face but Sadness in the Eyes: Emotion Recognition based on Facial Expressions and Eye Behaviors3
A comprehensive analysis of the importance of eye behaviors in ER, advancing the study on addressing the gap between FER and ER for more robust ER performance and specifically design a new EMERT architecture to concurrently enhance performance in both ER and FER by efficiently identifying and bridging the emotion gap between the two.
- Pluralistic Image Completion with Probabilistic Mixture-of-Experts3
A unified probabilistic graph model is introduced that represents the complex interactions in image completion and is formally established the effectiveness of the method and demonstrated it with comprehensive experiments.
- Original or Translated? On the Use of Parallel Data for Translation Quality Estimation3
This paper compares the impact of parallel data with different translation directions in QE data augmentation, and finds that using the source-original part of parallel corpus consistently outperforms its target-original counterpart.
- 3
- 3
- Relationship graph learning network for visual relationship detection3
A relationship graph learning network (RGLN) to explore the correlations among objects' relationships for visual relationship detection and results compared with the state-of-the-art demonstrate the competitiveness of RGLN.
- Contrastive knowledge embedding with discriminative self-weighted sampling2
The proposed CoDiSS framework can improve the performance of various KG embedding models, such as TransE, ComplEx, and HousE, enabling them to produce expressive KG embeddings.
- 2
- Focusing on pedestrians like human for clothes changing person re-identification2
This paper proposes a humanoid focus-inspired image augmentation (HFIA) method, which is intuitive image processing rather than a sophisticated network architecture designed to enhance local nuances of pedestrian images.
- 2
- 2
- From Answers to Rationales: Self-Aligning Multimodal Reasoning with Answer-Oriented Chain-of-Thought2
Experiments indicate that the SMART framework significantly improves various MLLMs, regardless of model architecture, parameter size, or pre-training dataset, and establishes an iterative generation-optimization method that continually enhances the model's reasoning skills.
- RGD-VNet: Raw, Generative, and Discriminant Views Network for Boosting Postoperative Complication Prediction2
A novel framework named Raw, Generative, and Discriminant Views Network (RGD-VNet) for PCP that can better mine the internal connections between labeled and unlabeled data is proposed.
- 2
- 1
- Re-Initialization Token Learning for Tool-Augmented Large Language Models1
A novel token learning method that aligns tool tokens with the existing word embedding space from the perspective of initialization, thereby enhancing model performance and effectively augments LLMs with tools through relevant tokens across diverse domains.
- 1
- 1
- 1
- 1
- A Comprehensive Data Preprocessing Framework towards Improving Internet Chinese Medical Data Quality1
- TD^2-Net: Toward Denoising and Debiasing for Dynamic Scene Graph Generation1
This paper proposes a denoising spatio-temporal transformer module that enhances object representation with robust contextual information and introduces an asymmetrical reweighting loss to relieve the issue of label bias.
- 1
- Careful Selection and Thoughtful Discarding: Graph Explicit Pooling Utilizing Discarded Nodes1
A novel Graph Explicit Pooling (GrePool) method is introduced, which selects nodes by explicitly leveraging the relationships between the nodes and final representation vectors crucial for classification, and enhances GrePool's performance without incurring additional computational costs.
- 1
- 1
- 1
- A Question Answering System for Unstructured Table Images1
This work presents a question answering system for unstructured table images that mainly consists of a table recognizer to recognize the tabular structure from an image and a table parser to generate the answer to a natural language question over the table.
- 1
- Graph Pattern Loss Based Diversified Attention Network For Cross-Modal Retrieval1
This paper proposes a Graph Pattern Loss based Diversified Attention Network (GPLDAN) for unsupervised cross-modal retrieval to deeply analyze correlations among representations and designs a novel graph pattern loss to explore the correlations among different representations.
- –
- –
- CERA: Conflict-Explicit Reflective Agent for Multimodal Emotion Reasoning–
Conflict-Explicit Reflective Agent (CERA), a training-free, conflict-aware, and language-driven agentic framework for MER, significantly outperforms state-of-the-art training-free methods in both recognition accuracy and conflict interpretability, providing an effective training-free solution for complex emotional reasoning.
- –
- –
- –
- –
- –
- –
- Towards Reliable Medical LLMs: Benchmarking and Enhancing Confidence Estimation of Large Language Models in Medical Consultation–
This work proposes the first benchmark for assessing confidence in multi-turn interaction during realistic medical consultations, and presents MedConf, an evidence-grounded linguistic self-assessment framework that constructs symptom profiles via retrieval-augmented generation, aligns patient information with supporting, missing, and contradictory relations, and aggregates them into an interpretable confidence estimate through weighted integration.
- –
- JOTGLNet: A Guided Learning Network with Joint Offset Tracking for Multiscale Deformation Monitoring–
- Semantic segmentation in power grid scenarios using scale-transforming transformer–
The approach’s novelty lies in a scale-transforming module (STM), which implements cost-effective aspect ratio adjustments, patch splitting, and patch combining, and which seamlessly integrates into existing transformer models to build STT models.
- Beyond Human Data: Aligning Multimodal Large Language Models by Iterative Self-Evolution–
This work proposes a novel multimodal self-evolution framework that empowers the model to autonomously generate high-quality questions and answers using only unannotated images and designs an answer self-enhancement technique to boost the discriminative power of answers.
- –
- –
- –
- –
- An Efficient Multi-prior Hybrid Approach for Consistent 3D Generation from Single Images–
Experimental results demonstrate the approach generates more consistent and realistic 3D assets than previous state-of-the-art methods in visual quality and quantitative metrics.
- –
- MCFR'22: 1st Workshop on Multimedia Computing towards Fashion Recommendation–
This workshop will showcase the innovative methodologies and ideas on new yet challenging research problems, including (not limited to) fashion recommendation, interactive fashionRecommendation, interactive garment retrieval, and outfit compatibility modeling.
- –
- Exploring High-Order Structure for Robust Graph Structure Learning–
This paper analyzes the adversarial attack on graphs from the perspective of feature smoothness which further contributes to an efficient new adversarial defensive algorithm for GNNs and proposes a novel algorithm that incorporates the high-order structural information into the graph structure learning.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.