Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks97 from public data
- 521
- All in One: Exploring Unified Video-Language Pre-Training258
This work introduces an end-to-end VLP model, namely all-in-one Transformer, that embeds raw video and textual signals into joint representations using a unified backbone architecture and introduces a novel and effective token rolling operation to encode temporal representations from video clips in a non-parametric manner.
- 155
- Removing the Background by Adding the Background: Towards Background Robust Self-supervised Video Representation Learning115
Self-supervised learning has shown great potentials in improving the video representation ability of deep neural networks by getting supervision from the data itself, but some of the current methods tend to cheat from the background, so this work proposes to remove the background impact by adding the background.
- A Simple Recipe towards Reducing Hallucination in Neural Surface Realisation92
This work proposes to integrate a language understanding module for data refinement with self-training iterations to effectively induce strong equivalence between the input data and the paired text to mitigate neural language generation systems often hallucinate contents.
- Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion73
This work proposes to decouple the scene and the motion (DSM) with two simple operations, so that the model attention towards the motion information is better paid and the impact of the scene is weakened while the temporal sensitivity of the network is further enhanced.
- Improving Entity Linking by Modeling Latent Entity Type Information72
This work proposes to inject latent entity type information into the entity embeddings based on pre-trained BERT, and integrates a BERT-based entity similarity score into the local context model of a state-of-the-art model to better capture latententity type information.
- Love Evolution Algorithm: a stimulus–value–role theory-inspired evolutionary algorithm for global optimization69
The efficiency of the LEA is validated through numerical experiments with CEC2017 benchmark functions, outperforming seven metaheuristic algorithms as evidenced by the Wilcoxon signed-rank test and the Friedman test.
- Cross-Batch Negative Sampling for Training Two-Tower Recommenders66
This work proposes a simple yet effective sampling strategy called Cross-Batch Negative Sampling (CBNS), which takes advantage of the encoded item embeddings from recent mini-batches to boost the model training.
- MambaVC: Learned Visual Compression with Selective State Spaces53
This work introduces MambaVC, a simple, strong and efficient compression network based on SSM, which achieves superior rate-distortion performance with lower computational and memory overheads and shows even greater improvements with high-resolution images, highlighting its potential and scalability in real-world applications.
- Position-Guided Text Prompt for Vision-Language Pre-Training50
This work proposes a novel Position-guided Text Prompt (PTP) paradigm to enhance the visual grounding ability of cross-modal models trained with VLP, and introduces PTP into several state-of-the-art VLP frameworks.
- Operation-guided Neural Networks for High Fidelity Data-To-Text Generation47
This paper proposes a framework called Operation-guided Attention-based sequence-to-sequence network (OpAtt), with a specifically designed gating mechanism as well as a quantization module for operation results to utilize information from pre-executed symbolic operations.
- GMMFormer: Gaussian-Mixture-Model Based Transformer for Efficient Partially Relevant Video Retrieval40
This paper proposes GMMFormer, a Gaussian-Mixture-Model based Transformer which models clip representations implicitly, and incorporates Gaussian-Mixture-Model constraints to focus each frame on its adjacent frames instead of the whole video.
- A Statistical Framework for Product Description Generation39
A statistical framework that generates accurate and fluent product description from product attributes by using learned knowledge to decide what to say and how to say for product description generation is presented.
- Enhancing Neural Data-To-Text Generation Models with External Background Knowledge37
This paper enhances neural data-to-text models with external knowledge in a simple but effective way to improve the fidelity of generated text by attends to relevant external knowledge, encoded as a temporary memory, and combines this knowledge with the context representation of data before generating words.
- What Makes for Good Visual Instructions? Synthesizing Complex Visual Reasoning Instructions for Visual Instruction Tuning33
It is found that instructions focusing on complex visual reasoning tasks are particularly effective in improving the performance of MLLMs, with results correlating to instruction complexity.
- 33
- Mention and Entity Description Co-Attention for Entity Disambiguation33
A type-aware co-attention model for entity disambiguation is proposed, which tries to identify the most discriminative words from mention contexts and most relevant sentences from corresponding entity descriptions simultaneously.
- Infer User Interests via Link Structure Regularization32
Both quantitative and qualitative experiments show that the proposed regularization framework based on the relation bipartite graph outperforms a few competitive baselines in learning user interests over a set of predefined topics and gives superior results compared to the baselines on retweet prediction and topical authority identification.
- Video-Text Pre-training with Learned Regions for Retrieval30
This work proposes a simple yet effective module for video-text representation learning, namely RegionLearner, which can take into account the structure of objects during pre-training on large-scale video- text pairs and is much more computationally efficient.
- 26
- Data2Text Studio: Automated Text Generation from Structured Data23
This work conducts experiments on RotoWire datasets for template extraction and text generation and shows that the Semi-HMMs model achieves improvements on both tasks.
- Adversarial open set domain adaptation via progressive selection of transferable target samples22
A novel method namely Thresholded Domain Adversarial Network (ThDAN), which progressively selects transferable target samples for distribution alignment and outperforms state-of-the-art domain adaptation and open set recognition approaches on benchmarks.
- 21
- Escape after love: Philoponella prominens optimizer and its application to 3D path planning20
A novel metaheuristic inspired by the special mating behavior of the P. prominens, and simulates escape, sexual cannibalism and predation behaviors of males after mating, which demonstrates PPO's competitiveness against six state-of-the-art metaheuristics.
- De Novo Generation and Identification of Novel Compounds with Drug Efficacy Based on Machine Learning20
The experimental results reveal that DTLS (Deep Transfer Learning‐based Strategy) can not only realize the generation and identification of novel compounds with drug efficacy but also has the advantage of identifying compounds by focusing on protein targets to facilitate the mechanism study.
- Enhancing Visual Grounding in Vision-Language Pre-Training With Position-Guided Text Prompts19
A novel Position-guided Text Prompt (PTP) paradigm is introduced to bolster the visual grounding abilities of cross-modal models trained with VLP, and integrates the seconda-order relationships between objects to further enhance the visual grounding capabilities of the proposed PTP paradigm.
- 19
- 18
- 18
- GMMFormer v2: An Uncertainty-aware Framework for Partially Relevant Video Retrieval15
GMMFormer v2, an uncertainty-aware framework for PRVR, is presented, demonstrating remarkable improvement of GMMFormer v2 compared to the past SOTA competitor and the versatility of uncertainty-aware text-clip matching for PRVR.
- CanvasEmb: Learning Layout Representation with Large-scale Pre-training for Graphic Design15
This paper proposes CanvasEmb (Canvas Embedding), which pre-trains deep representations from unlabeled graphic designs by jointly conditioning on all the context elements in a canvas, with a multi-dimensional feature encoder and aMulti-task learning objective.
- Weakly Supervised Deep Hyperspherical Quantization for Image Retrieval14
This work proposes Weakly-Supervised Deep Hyperspherical Quantization (WSDHQ), which is the first work to learn deep quantization from weakly tagged images and uses word embeddings to represent the tags and enhance their semantic information based on a tag correlation graph.
- MUP: Multi-granularity Unified Perception for Panoramic Activity Recognition13
A novel Multi-granularity Unified Perception (MUP) framework that perceives different granularity behaviors universally to explore the co-occurrence motion pattern via the same parameters in an end-to-end fashion is proposed.
- 13
- 13
- Learning Latent Semantic Annotations for Grounding Natural Language to Structured Data12
This paper attempt at learning explicit latent semantic annotations from paired structured tables and texts, establishing correspondences between various types of values and texts with an adapted semi-hidden Markov model.
- 11
- Hierarchical Mixed Neural Network for Joint Representation Learning of Social-Attribute Network10
In real-world networks, text or categorical attributes are usually associated with nodes, providing another description for networks in a different perspective, and network representation learning methods designed for homogeneous network, which only consider topological properties of networks.
- 9
- 9
- 9
- Suppressing Static Visual Cues via Normalizing Flows for Self-Supervised Video Representation Learning9
This work proposes a novel method to suppress static visual cues (SSVC) based on probabilistic analysis for self-supervised video representation learning that outperforms the state of the art when only single RGB modality is used for pre-training.
- 9
- Aggregated Semantic Matching for Short Text Entity Linking9
A novel neural network framework, Aggregated Semantic Matching (ASM), in which two different aspects of semantic information between the local context and the candidate entity are captured via representation-based and interaction-based neural semantic matching models, and then two matching signals work jointly for disambiguation with a rank aggregation mechanism.
- Originator or propagator?9
This paper proposes a method inspired from Social Role Theory (SRT), which assumes that a user behaves differently with different roles in the generation process of Twitter content, and considers the two most distinctive social roles on Twitter: originator and propagator.
- AutoSSVH: Exploring Automated Frame Sampling for Efficient Self-Supervised Video Hashing8
This work proposes a new framework, termed AutoSSVH, that employs adversarial frame sampling with hash-based contrastive learning and introduces a hash component voting strategy and a point-to-set hash-based contrastive objective, which help capture complex inter-video semantic relationships in the Hamming space and improve the discriminability of learned hash codes.
- 8
- 7
- 7
- 5
- Towards Scalable Semantic Representation for Recommendation5
This paper proposes Mixture-of-Codes, which first constructs multiple independent codebooks for LLM representation in the indexing stage, and then utilizes the Semantic Representation along with a fusion module for the downstream recommendation stage, leading to the best scale-up performance in recommendations.
- Hypergraph-Guided Disentangled Spectrum Transformer Networks for Near-Infrared Facial Expression Recognition5
This paper gives the first attempt at deep NIR facial expression recognition and proposes a novel method called near-infrared facial expression transformer (NFER-Former), a Self-Attention Orthogonal Decomposition mechanism that disentangles the expression information and spectrum information from the input image, so that the expression features can be extracted without the interference of spectrum variation.
- An Encoder with non-Sequential Dependency for Neural Data-to-Text Generation5
This work proposes to adopt the self-attention mechanism to learn dependencies between arbitrary input records and shows the proposed method achieves comparable results and remains stable under random shuffling over input data.
- 4
- Bayesian Decision Model Based on Probabilistic Rough Set with Variable Precision4
The basic process of minimum risk of Bayesian decision is discussed, the probabilistic rough set model with variable precision is set up, and the model could decrease the error risk of the Bayesian decision.
- 4
- Kernel extreme learning machine optimized by arithmetic optimization algorithm3
This work introduces the Kernel Extreme Learning Machine optimized by the Arithmetic Optimization Algorithm (AOA-KELM) in an effort to significantly improve the efficiency of the Kernel Extreme Learning Machine.
- DFFA-Net: A Differential Convolutional Neural Network for Underwater Optical Image Dehazing3
DFFA-Net is obtained by deeply analyzing the factors that affect the quality of underwater images and combining the underwater light propagation characteristics, and introduces a channel differential module that captures the mutual information between the green and blue channels with respect to the red channel.
- 3
- 3
- 3
- 3
- 2
- 2
- 2
- 2
- Learning Spatio-temporal Representation by Channel Aliasing Video Perception2
This paper proposes a novel pretext task namely Channel Aliasing Video Perception (CAVP) for self-supervised video representation learning, to recognize the number of different motion flows within a channel aliasing video for perception of discriminative motion cues.
- 2
- Beyond Word for Word: Fact Guided Training for Neural Data-to-Document Generation2
A Fact Guided Training (FGT) model is proposed to improve both content selection and surface realization by leveraging an information extraction (IE) system and experimental results show the proposed model outperforms neural encoder-decoder models in both automatic and human evaluations.
- Using Lexical and Thematic Knowledge for Name Disambiguation2
A novel approach to disambiguate names based on two different types of semantic information: lexical and thematic, using translation-based language models to resolve the synonymy problem in every word match and a topic-based ranking function to capture rich thematic contexts for names.
- DOA estimation algorithm of MIMO system2
Simulation results indicate that the CSM is a high resolution and robustness algorithm for resolving multiple incoherent and coherent signals that are incident on an array of antenna.
- Flexible Strain Sensor Enabled by Back Propagation Neural Network for Gesture Recognition1
A flexible strain sensor is developed based on microfluidics using a back propagation (BP) neural network for highly accurate gesture recognition and has broad application potential in biomedicine, aerospace, and other fields.
- 1
- 1
- 1
- Spatial-Temporal Bottom-Up Top-Down Attention Model for Action Recognition1
Spatial-temporal Bottom-up Top-down Attention Module (STBTA) can be plugged into any feedforward network architectures and is end-to-end trainable along with CNN.
- Evolution Analysis of Adaptive Distance-Preserving Level Set for Image Segmentation Based on Depth Learning1
In the future, the focus of the work will be on improving the generalization ability of the model so that the model can perform well in a variety of tasks.
- 1
- –
- –
- Research on multimodal target tracking system based on forest environment–
A multi-modal target tracking system designed for the complex forest environment that improves detection accuracy but also sets a new technical path for target tracking in complex environments, providing a valuable reference for future studies and practical applications.
- –
- –
- –
- –
- –
- Learning Semantic Correspondences from Noisy Data-text Pairs by Local-to-Global Alignments–
This work proposes a two-stage local-to-global alignment (L2GA) framework that can work jointly to learn semantic correspondences in the same framework and shows that the proposed method can be generalized to both restaurant and computer domains and improve the alignment accuracy.
- –
- –
- A Novel Spectral Matching Algorithm to Application Environment Fitness Evaluation Method–
The accuracy of the evaluation method is high, the evaluation rules basically meet the requirements, and the fitness evaluation result obtained is obtained by fuzzy comprehensive evaluation method.
- –
- Adaptive Routing Algorithm for Multi-hop Virtual Cellular Network–
A table-driven adaptive routing algorithm applied for port addition, port removal and link/port failure is proposed, and the transmit power efficiency is evaluated by computer simulation.
- A Joint Frequency-Domain Equalization (Fde) and Antenna Diversity Combining Method for the Multipath Fading in the Frequency-Selective Channel–
A theoretical foundation is built to evaluate the downlink performance of an MC-CDMA cellular system with site diversity operation with and receive antenna diversity combining and results from theoretical computation are compared to the results from computer simulation.
- Research on measurement regions for Articulated Arm Coordinate Measuring Machines based on DBSCAN–
The experiment showed that the measurement accuracy was greatly influenced by the measurement regions, and precision could be improved by measuring in high precision regions.
- –
- –
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.