Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks107 from public data
- 260
- Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark163
E evaluation of state-of-the-art MLLMs on EMMA reveals significant limitations in handling complex multimodal and multi-step reasoning tasks, even with advanced techniques like Chain-of-Thought prompting and test-time compute scaling underperforming, underscore the need for improved multimodal architectures and training paradigms to close the gap between human and model reasoning in multimodality.
- Distilling Knowledge Learned in BERT for Text Generation147
A novel approach, Conditional Masked Language Modeling (C-MLM), is presented to enable the finetuning of BERT on target generation tasks, which significantly outperforms strong Transformer baselines on multiple language generation tasks such as machine translation and text summarization.
- Look across Elapse: Disentangled Representation Learning and Photorealistic Cross-Age Face Synthesis for Age-Invariant Face Recognition104
This work proposes a deep Age-Invariant Model (AIM) for face recognition in the wild with three distinct novelties, and proposes a new large-scale Cross-Age Face Recognition (CAFR) benchmark dataset to facilitate existing efforts and push the frontiers of age-invariant face recognition research.
- Peptide-drug conjugates: A new paradigm for targeted cancer therapy88
Based on the structure and function of PDCs, the latest research progress on PDCs from the aspects of cancer cell-targeting peptide (CTP) selection, pharmacokinetic characteristics, stability regulation and so on were systematically reviewed, hoping to highlight the current problems and future development directions of PDCs.
- 83
- 3D-Aided Deep Pose-Invariant Face Recognition83
A 3D-Aided Deep Pose-Invariant Face Recognition Model (3D-PIM), which automatically recovers realistic frontal faces from arbitrary poses through a 3D face model in a novel way, and demonstrates superiority of the proposed model over state-of-the-arts.
- What is the difference between augmented reality and 2D navigation electronic maps in pedestrian wayfinding?75
Augmented reality (AR) navigation aids have become widely used in pedestrian navigation, yet few studies have verified their usability from the perspective of human spatial cognition, such as visual attention, cognitive processing, and spatial memory.
- TileLang: A Composable Tiled Programming Model for AI Systems72
TileLang is presented, a generalized tiled programming model for more efficient AI Kernel programming that decouples scheduling space from dataflow, and encapsulated them as a set of customization annotations and primitives, allowing users to focus on the kernel's data-flow itself, while leaving most other optimizations to compilers.
- 66
- Interleaving Reasoning for Better Text-to-Image Generation40
This work introduces Interleaving Reasoning Generation (IRG), a framework that alternates between text-based thinking and image synthesis: the model first produces a text-based thinking to guide an initial image, then reflects on the result to refine fine-grained details, visual quality, and aesthetics while preserving semantics.
- 40
- Rethinking the Video Sampling and Reasoning Strategies for Temporal Sentence Grounding39
A novel Siamese Sampling and Reasoning Network (SSRN) for TSG is proposed, which introduces a siamese sampling mechanism to generate additional contextual frames to enrich and refine the new boundaries.
- 39
- Diversity-Incentivized Exploration for Versatile Reasoning33
DIVER is proposed, an innovative framework that highlights the pivotal role of global sequence-level diversity to incentivize deep exploration for versatile reasoning and introduces global diversity incentives as an intrinsic reward to promote deep exploration in a semantically structured space.
- The recovery of the microbial community after plaque removal depends on periodontal health status33
It was found that the more severe the periodontal disease condition, the more lagged the recovery of plaque microorganisms to their original state after cleaning; this reflected a higher degree of plaque development in periodontitis samples.
- QuantMoE-Bench: Examining Post-Training Quantization for Mixture-of-Experts30
This paper explores several MoE structure-aware quantization heuristics, ranging from coarse to fine granularity, from MoE block to individual linear weight, and reveals critical principles: different MoE structures require varying numbers of weight bits for effective and efficient quantization.
- Look, Compare, Decide: Alleviating Hallucination in Large Vision-Language Models via Multi-View Multi-Path Reasoning29
A training-free framework that aims to reduce hallucinations in Large Vision-Language Models by making the most of the innate capabilities of the LVLMs via multi-path reasoning for each information view to quantify and aggregate the certainty scores for each potential answer among multiple decoding paths and finally decide the output answer.
- 27
- A Semantic Context-Based Model for MobileWeb Services Access Control22
A context ontology to represent contextual information and employ it in the inference engine of the SCBAC model is proposed, which uses semantic web rule language (SWRL) to form policy rule and infer those rules by JESS inference engine.
- 21
- 21
- 21
- 21
- HASA-Net: A Non-Intrusive Hearing-Aid Speech Assessment Network20
This work proposes a DNN-based hearing aid speech assessment network (HASA-Net), formed by a bidirectional long short-term memory (BLSTM) model, to predict speech quality and intelligibility scores simultaneously according to input speech signals and specified hearing-loss patterns.
- How do voice-assisted digital maps influence human wayfinding in pedestrian navigation?19
Overall, the use of voice-assisted digital maps saves time but may reduce pedestrians’ spatial knowledge acquisition, and the scientific understanding of the multimedia navigation mode in shaping human spatial ability is deepen.
- 17
- Trusted autonomic service cooperation model and application development framework15
This paper proposes a model of Norm-Governed and Policy-Driven autonomic service cooperation (NGPD), which can overcome the limitation introduced by non-autonomic service cooperation, and thus make the autonomy and change-response ability of service cooperation exhibit the advantages of robustness and intelligence which cannot be reached by traditional service cooperation techniques.
- Discovery of genistein derivatives as potential SARS-CoV-2 main protease inhibitors by virtual screening, molecular dynamics simulations and ADMET analysis13
In silico work results show that the preferred compound 11, which exhibited strong binding to SARS-CoV-2 main protease, acceptable drug-like properties, protease inhibitory activity and ADMET properties, has great promise for further research as a potential therapeutic agent against COVID-19.
- 13
- 11
- Attribute assisted teacher-critical training strategies for image captioning11
A teacher model is introduced that serves as a bridge between the ground-truth caption and the caption model by generating some easier-to-learn word proposals as soft targets to facilitate better learning processes for the Caption model.
- 10
- SI-Net: spatial interaction network for deepfake detection10
Spatial Interaction Network (SI-Net) is proposed, a deepfake detection method to mine potential complementary and co-occurrent features between local texture and global context concurrently, concurrently.
- The interactive fuzzy linguistic term set and its application in multi-attribute decision making10
The properties of the interactive fuzzy linguistic term set and its advantages of improving the consistency of decision information are discussed and some numerical examples are given to illustrate its application in dealing with the interactive information in multi-attribute decision making problems.
- Improving generative adversarial network inversion via fine-tuning GAN encoders9
A self-supervised method to pre-train and fine-tune GAN encoders using synthesized images and fine-tune the pre-trained GAN encoder for inverting real images.
- 9
- 8
- Fast 2-step regularization on style optimization for real face morphing7
This work proposes a novel method to morph real faces based on StyleGAN by adding regularization to style optimization, and uses learned directions under different layer representations to handle real face morphing.
- 7
- Reducing infrequent-token perplexity via variational corpora7
A generic RNN model is adapted and it is shown that, with variational training corpora and epoch unfolding, the model improves its performance for the task of URL insertion suggestions.
- A Context-Aware Semantic-Based Access Control Model for Mobile Web Services7
The proposed model adopts a context-centric policy method, and grants permissions to users according to current context information and allows high-level description and reasoning about contexts and policies.
- 6
- 6
- 6
- 6
- Self-Supervised Animation Synthesis Through Adversarial Training6
A novel deep generative model based on self-supervised learning and adversarial training, the model can find labeling rules and mark them without origin sample labels based on the automatically labels learning and shows that an appropriate number of the pseudo-label dimensions can better characterize animation features.
- 6
- 5
- Mitochondrial transplantation—a novel therapeutic strategy for erectile dysfunction: a narrative review3
Although clinical translation faces challenges such as scalable production, dosage standardization, and long-term immunocompatibility, MT holds promise as a mechanism-based therapy for refractory ED, offering new insights beyond conventional hemodynamic approaches.
- A VVC Dependent Quantization Optimization Based on the Parallel Viterbi Algorithm and Its FPGA Implementation3
The proposed algorithm reduces the quantization cycle by an average of 56.96% compared to VVC’s reference platform VTM, and the hardware implementation meets the quantization requirements for 1080P@60Hz video hardware encoding on the AMD FPGA development platform.
- Perceptual Characteristics Based Multi-objective Model for Speech Enhancement3
Experimental results show the proposed algorithm evidently outperforms the original SE model in all objective scores, including speech quality, speech intelligibility and signal distortion.
- A Semantic-Aware Context-Based Access Control Framework for Mobile Web Services3
A context ontology is proposed to represent contextual information and employ it in the inference engine and extending XACML to incorporate the knowledge base by combining semantic web technologies with context-based access control mechanism.
- 2
- Research on Challenges and Strategies of Students' Adaptive Learning within AI2
It's found that adaptive learning is mainly experiencing the challenges involving how to avoid data bias and protect personal privacy, making continuous evaluations of users, lack of emotion monitoring and technological support for both students and teachers and the technology integration and inequity.
- Low-Rank Active Learning for Generating Speech-Drive Human Face Animation2
A novel facial animation technique that can intelligently generates real human face animation videos by leveraging an emotional speech is introduced and a low-rank active learning paradigm for discovering multiple key facial frames that can best represent the above phoneme and emotion pairs in the feature subspace is proposed.
- Research on Chinese Word Segmentation Algorithm Based on Special Identifiers2
This paper provides aword segmentation method based on special identifiers, and realizes a word segmentation system by combining the special identifier set with the modified two-character dictionary structure before it carries out the comparison test for that system and other word segmentations systems by SOUGOU training corpus’s test text.
- The dynamic elimination effect of local adjuvant therapy on the periodontal microbiota1
Minocycline hydrochloride local adjunctive therapy outperforms semiconductor laser local adjunctive therapy in terms of eliminating periodontitis core bacteria, improving the microbial community structure, and enhancing clinical indicators.
- Towards Interpretable Face Morphing via Unsupervised Learning of Layer‐Wise and Local Features1
An improved unsupervised approach that leverages contrastive learning and K‐means clustering to learn both layer‐wise and local features (LLF) in the latent space of StyleGAN, offering a promising solution for enhanced unsupervised face morphing applications.
- 1
- 1
- 1
- 1
- 1
- 1
- Pairwise FastText Classifier for Entity Disambiguation1
PFC is devised, an efficient embedding-based text classifier, and used for entity disambiguation and a method to bootstrap the training set deterministically without human labelling and at no financial cost is created.
- Research on Project Management system of Large Complex Construction Projects Based on cloudDB1
Important functions and implementation method of key technologies are advocated in this paper and the structure chart and functional diagram of large and complex project management systems based on cloud databases are presented.
- 1
- 1
- 1
- Humanoid Recognition of Handwritten Chinese Character Based on Procedure Neural Networks1
A kind of characteristic knowledge data-base table of handwritten Chinese characters is constructed, and the characters are recognized by comparing and judging and this method is proved to have ability of "cognizing".
- –
- –
- –
- Chinese ModernBERT with Whole-Word Masking–
Chinese ModernBERT is introduced, a from-scratch Chinese encoder that couples a hardware-aware 32k BPE vocabulary tailored to frequent Chinese affixes/compounds, lowering the embedding budget and releasing tokenizer and weights to facilitate reproducible research.
- –
- –
- Developing the Influencer Marketing Process based on the Data-Driven Service Design–
This research provided ideas and a practical case for future influencer marketing and social media-related service design research and proposed a social media-driven service design methodology.
- Ask Me Any Type: Type Inference Plugin for Partial Code on the Web and in the Integrated Development Environment–
Ask Me Any Type (AMAT), a type of inference plugin embedded in web browsers and integrated development environment (IDE), uses a cloze-style fill-in-the-blank strategy for type inference and outperforms state-of-the-art tools such as COSTER and SNR.
- Analytic Convolutional Layer: A Step to Analytic Neural Network–
The Analytic Convolutional Layer (ACL), an innovative model-driven convolutional layer, which is a mosaic of analytical convolution kernels (ACKs) and traditional convolution kernels (ACKs) and traditional convolution kernels, is introduced.
- –
- Research on the Development Path of Online Referral based on the Background of Smart Medicine–
The problems and challenges faced in the current development of smart medicine in China are summarized, the current situation of referral under the hierarchical diagnosis and treatment system in China is focused on, and the problems of complicated procedures and difficult transfer to the referral mode under the qualification line are analyzed.
- –
- A Novel Approach of Automatic Judgment for Subjective Questions–
A novel approach that develops a similarity calculation model, which combines the Word2Vec model and the keyword weight TF-IDF algorithm to automatically score subjective questions is proposed.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Subjective carnival——a study on the subject of art in network–
The subject of art in network is a person who takes internet as platform and consciously engages in creating, communicating, evaluating in artistic fields and embodies some characteristics of the "carnival" theory of Mikhail Bakhtin.
- –
- –
- –
- –
- –
- –
- –
- –
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 2026-10-15. 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.