Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks181 from public data
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing188
An in-depth probing analysis is conducted and it is demonstrated that cross-attention maps in Stable Diffusion often contain object attribution information, which can result in editing failures, and proposes a simplified, yet more stable and efficient, tuning-free procedure that modifies only the self-attention maps of specified attention layers during the denoising process.
- A retrospective of knowledge graphs158
This paper discusses in detail about the process of building knowledge graphs, and survey state-of-the-art techniques for automatic knowledge graph checking and expansion via logical inferring and reasoning.
- Development and evaluation of a Web 2.0 annotation system as a learning tool in an e-learning environment140
The evaluation results showed that, with factor analysis, interactivity, usefulness, helpfulness, and willingness for future use were categorized to represent the perceptions of MyNote and the factors of interactivity and helpfulness were statistically significant to predict the future use of My note.
- Video Generation From Single Semantic Label Map111
A cVAE for predicting optical flow is employed as a beneficial intermediate step to generate a video sequence conditioned on the initial single frame, and a semantic label map is integrated into the flow prediction module to achieve major improvements in the image-to-video generation process.
- 94
- A Customized Text Sanitization Mechanism with Differential Privacy93
A novel Customized Text (CusText) sanitization mechanism based on the original $\epsilon$-differential privacy (DP) definition, which is compatible with any similarity measure, and achieves a better trade-off between privacy and utility than existing mechanisms.
- A Short Survey on Taxonomy Learning from Text Corpora: Issues, Resources and Recent Advances71
Recent advances on taxonomy construction from free texts are overviewed, reorganizing relevant subtasks into a complete framework for evaluation and discussing challenges for future research.
- Meta-KD: A Meta Knowledge Distillation Framework for Language Model Compression across Domains70
This work proposes a Meta-Knowledge Distillation (Meta-KD) framework to build a meta-teacher model that captures transferable knowledge across domains and passes such knowledge to students and demonstrates the effectiveness and superiority of this framework on public multi-domain NLP tasks.
- TransPrompt: Towards an Automatic Transferable Prompting Framework for Few-shot Text Classification66
In TransPrompt, a transferable prompting framework for few-shot learning across similar NLP tasks, a multi-task meta-knowledge acquisition procedure is employed to train a meta-learner that captures cross-task transferable knowledge.
- 62
- SMedBERT: A Knowledge-Enhanced Pre-trained Language Model with Structured Semantics for Medical Text Mining61
In SMedBERT, a medical PLM trained on large-scale medical corpora, incorporating deep structured semantic knowledge from neighbours of linked-entity, the mention-neighbour hybrid attention is proposed to learn heterogeneous-entity information, which infuses the semantic representations of entity types into the homogeneous neighbouring entity structure.
- Making Pre-trained Language Models End-to-end Few-shot Learners with Contrastive Prompt Tuning44
CP-Tuning is presented, an end-to-end Contrastive Prompt Tuning framework for fine-tuning PLMs without any manual engineering of task-specific prompts and verbalizers, and is integrated with the task-invariant continuous prompt encoding technique with fully trainable prompt parameters.
- HiCLRE: A Hierarchical Contrastive Learning Framework for Distantly Supervised Relation Extraction44
A three-level hierarchical learning framework to interact with cross levels is proposed, generating the de-noising context-aware representations via adapting the existing multi-head self-attention, named Multi-Granularity Recontextualization.
- DKPLM: Decomposable Knowledge-Enhanced Pre-trained Language Model for Natural Language Understanding44
A novel KEPLM named DKPLM is proposed that decomposes knowledge injection process of the pre-trained language models in pre-training, fine-tuning and inference stages, which facilitates the applications of KEPLMs in real-world scenarios.
- Improving Clinical Named Entity Recognition with Global Neural Attention40
The authors' model outperforms previous state-of-the-art method with less labeled data and no feature engineering, and is evaluated on 2010 i2b2/VA datasets to verify the effectiveness of leveraging global information and transfer strategy.
- 35
- 35
- 35
- From Dense to Sparse: Contrastive Pruning for Better Pre-trained Language Model Compression34
ContrAstive Pruning (CAP) is proposed, designed as a general framework, compatible with both structured and unstructured pruning, and enables the pruned model to learn from the pre-trained model for task-agnostic knowledge, and fine-tuned model fortask-specific knowledge.
- MeLL: Large-scale Extensible User Intent Classification for Dialogue Systems with Meta Lifelong Learning34
The Meta Lifelong Learning (MeLL) framework is introduced, a BERT-based text encoder is employed to learn robust text representations across tasks, which is slowly updated for lifelong learning and improves the performance compared with strong baselines and also reduces the number of total parameters.
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding33
This paper introduces DocLayLLM, an efficient multi-modal extension of LLMs specifically designed for TDU, and deeply considered the role of chain-of-thought (CoT) and innovatively proposed the techniques of CoT Pre-training and CoT Annealing.
- MCL: Multi-Granularity Contrastive Learning Framework for Chinese NER28
This paper proposes a novel Multi-Granularity Contrastive Learning framework (MCL), that aims to optimize the inter-granularity distribution distance and emphasize the critical matched words in the lexicon and obtains state-of-the-art results while considering model efficiency.
- 26
- 24
- Meta Fine-Tuning Neural Language Models for Multi-Domain Text Mining24
An effective learning procedure named Meta Fine-Tuning (MFT), served as a meta-learner to solve a group of similar NLP tasks for neural language models, and is implemented upon BERT to solve several multi-domain text mining tasks.
- 23
- On the Trustworthiness Landscape of State-of-the-art Generative Models: A Survey and Outlook21
An extensive survey is developed that outlines the trustworthiness of large generative models and practical recommendations are provided that provide practical recommendations and identify promising research directions for generative AI, ultimately promoting the trustworthiness of these models and benefiting society as a whole.
- EasyTransfer20
The EasyTransfer platform is designed to develop deep TL algorithms for NLP applications, backended with a high-performance and scalable engine for efficient training and inference, and also integrates comprehensive deepTL algorithms, to make the development of industrial-scale TL applications easier.
- A Family of Fuzzy Orthogonal Projection Models for Monolingual and Cross-lingual Hypernymy Prediction20
This work presents a family of fuzzy orthogonal projection models for both monolingual and cross-lingual hypernymy prediction, and establishes distributional fuzzy mappings from embeddings of a term to those of its hypernyms and non-hypernyms, which consider the complicated linguistic regularities of these relations.
- Transductive Non-linear Learning for Chinese Hypernym Prediction20
This paper presents a transductive learning approach to establish mappings from entities to hypernyms in the embedding space directly, which combines linear and non-linear embedding projection models, with the capacity of encoding arbitrary language-specific rules.
- User Generated Content Oriented Chinese Taxonomy Construction20
This paper takes Chinese Wikipedia as the data source, develops methods to extract classes and their relations mined from user tagged categories, and builds up the taxonomy using a bottom-up strategy, and shows that the constructed Chinese taxonomy achieves better results in both quality and quantity.
- AGREE: Aligning Cross-Modal Entities for Image-Text Retrieval Upon Vision-Language Pre-trained Models19
This paper proposes a lightweight and practical approach to align cross-modal entities for image-text retrieval upon VLP models only at the fine-tuning and re-ranking stages, which achieves state-of-the-art results in nearly all settings.
- BiRRE: Learning Bidirectional Residual Relation Embeddings for Supervised Hypernymy Detection19
A representation learning framework named Bidirectional Residual Relation Embeddings (BiRRE), which models the possibility of a term being mapped to another in the embedding space by hypernymy relations, and which outperforms strong baselines over various evaluation frameworks.
- 18
- 18
- 17
- SphereRE: Distinguishing Lexical Relations with Hyperspherical Relation Embeddings17
This work presents a neural representation learning model to distinguish lexical relations among term pairs based on Hyperspherical Relation Embeddings (SphereRE), which outperforms state-of-the-arts benchmarks.
- 16
- KEML: A Knowledge-Enriched Meta-Learning Framework for Lexical Relation Classification16
A neural classifier integrated with special relation recognition cells is proposed in order to combine meta-learning over the auxiliary task distribution and supervised learning for LRC, and Experiments over multiple datasets show KEML outperforms state-of-the-art methods.
- 16
- Predicting hypernym–hyponym relations for Chinese taxonomy learning16
This paper introduces a weakly supervised framework to extract Chinese is-a relations from user-generated categories that employs piecewise linear projection models trained on an existing Chinese taxonomy built from Wikipedia and an iterative learning algorithm to update model parameters incrementally.
- 15
- Chinese Hypernym-Hyponym Extraction from User Generated Categories15
A pattern-based relation selection method is proposed to prevent “semantic drift” in the learning process using bi-criteria optimization and experimental results illustrate that the proposed approach outperforms state-of-the-art methods.
- Oblivious Transfer via Lossy Encryption from Lattice‐Based Cryptography14
This paper first design a multibit lossy encryption under the decisional learning with errors (LWE) assumption and then design a new variant of UC-secure OT protocol for authenticated protocol via lossyryption scheme.
- Error Link Detection and Correction in Wikipedia14
This paper introduces an efficient method to generate candidate error links based on iterative ranking in an Anchor Text Semantic Network, which greatly reduces the problem space and proposes algorithms to detect and correct error links.
- Copper-catalyzed synthesis of 1,2,4-trisubstituted pyrroles via cascade reactions of aryloxy-enynes with amines13
A highly efficient copper-catalyzed cascade reaction of aryloxy-enynes with amines under mild reaction conditions has been developed and offers rapid access to 1,2,4-trisubstituted pyrroles in good to excellent yields in a regioselective manner.
- 12
- Guardian: Guarding against Gradient Leakage with Provable Defense for Federated Learning12
This paper proposes a novel yet effective defense framework named guarding against gradient leakage (Guardian) that produces transformed gradients by jointly optimizing two theoretically-derived metrics associated with gradients for performance maintenance and privacy protection and designs an ingenious initialization strategy for faster generation of transformed gradients.
- 12
- Event phase oriented news summarization12
In this approach, it is assumed that a summary contains multiple timelines, each corresponding to an event phase, and a structural clustering algorithm EPCluster is designed to separate news articles into several groups corresponding to event phases.
- ConaCLIP: Exploring Distillation of Fully-Connected Knowledge Interaction Graph for Lightweight Text-Image Retrieval11
The fully-Connected knowledge interaction graph (Coca) technique for cross-modal pre-training distillation is proposed and the resulting CocaCLIP achieves SOTA performances on the widely-used Flickr30K and MSCOCO benchmarks under the lightweight setting.
- Cross-domain Knowledge Distillation for Retrieval-based Question Answering Systems11
This work provides a new perspective on the potential of the teacher-student paradigm facilitating cross-domain transfer learning, where the teacher and student tasks belong to heterogeneous domains, with the goal to improve the student model’s performance in the target domain.
- DKGBuilder: An Architecture for Building a Domain Knowledge Graph from Scratch11
A domain knowledge graph construction system DKGBuilder is presented, which utilizes a template-based approach to extract seed knowledge from semi-structured data and employs an is-a relation classifier to learn a domain taxonomy using a bottom-up strategy.
- 10
- 9
- 9
- Resizing codebook of vector quantization without retraining9
Hyperbolic embeddings are leveraged to enhance codebook vectors with the co-occurrence information and reorder the enhanced codebook by the Hilbert curve so that the codebook of vector quantization can be resizeed for lower computation load or better reconstruction quality.
- 9
- TaxiHailer: A Situation-Specific Taxi Pick-Up Points Recommendation System9
This demonstration presents TaxiHailer, a situation-specific recommendation system for passengers who are eager to find a taxi, given a query with departure point, destination and time, which recommends pick-up points within a specified distance and ranked by potential waiting time.
- 8
- 8
- 8
- Meta Distant Transfer Learning for Pre-trained Language Models8
Inspired by meta-learning, the Meta Distant Transfer Learning (Meta-DTL) framework is proposed to learn the cross-task knowledge for PLM-based methods and is found to be highly effective when very few data is available for the target task.
- HORNET8
A novel KEPLM named HORNET is proposed, which integrates Heterogeneous knowledge from various structured and unstructured sources into the Roberta NETwork and hence takes full advantage of both linguistic and factual knowledge simultaneously, and design a hybrid attention heterogeneous graph convolution network (HaHGCN).
- 8
- 8
- 6
- Understanding Attention Mechanism in Video Diffusion Models6
An in-depth perturbation analysis of the spatial and temporal attention blocks of T2V models using an information-theoretic approach indicates that temporal and spatial attention maps affect not only the timing and layout of the videos but also the complexity of spatiotemporal elements and the aesthetic quality of the synthesized videos.
- 6
- 6
- 6
- 6
- 6
- 5
- 5
- 5
- Building Natural Language Processing Applications with EasyNLP5
The EasyNLP toolkit is designed for building PTM-based NLP applications with ease, which supports a comprehensive suite of NLP algorithms and is suitable for meeting the inference requirements in industry.
- 5
- Knowledge-Empowered Representation Learning for Chinese Medical Reading Comprehension: Task, Model and Resources5
The Chinese medical BERT model for the task (CMedBERT), which fuses medical knowledge into pre-trained language models by the dynamic fusion mechanism of heterogeneous features and the multi-task learning strategy, is proposed.
- 5
- 5
- 4
- 4
- 4
- Accelerating BERT inference with GPU-efficient exit prediction4
A new BERT inference method with GPU-Efficient Exit Prediction (GEEP), which leverages the shared exit loss to simplify the training process of FastBERT from two steps into only one step and makes the teacher classifier more knowledgeable by feeding diverse Transformer outputs to the teacherclassifier.
- FastBlend: a Powerful Model-Free Toolkit Making Video Stylization Easier4
A powerful model-free toolkit called FastBlend is proposed to address the consistency problem for video processing, based on a patch matching algorithm, and it outperforms existing methods for video deflickering and video synthesis.
- When Few-Shot Learning Meets Large-Scale Knowledge-Enhanced Pre-training: Alibaba at FewCLUE4
This paper presents the solution to FewCLUE tasks by means of large-scale knowledge-enhanced pre-training over massive texts and knowledge triples, together with a new few-shot learning algorithm for downstream tasks.
- Learning to Expand4
This work proposes a reinforced selector to extract useful PRF terms to enhance response candidates and a BERT-based response ranker to rank the PRF-enhanced responses and proves the superiority of the reinforced PRF term selector compared with other potential soft or hard selection methods.
- 4
- 4
- 4
- 3
- 3
- 3
- 3
- 3
- Exploiting Pre-Trained Models and Low-Frequency Preference for Cost-Effective Transfer-based Attack3
This work proposes a novel frequency perspective to study the transferability of cross-model transferability and identifies two factors that impair the transferability: an unchangeable intrinsic difference term along with a controllable perturbation-related term and suggests employing pre-trained models as proxy models, which are freely available.
- 3
- 3
- Rapid Diffusion: Building Domain-Specific Text-to-Image Synthesizers with Fast Inference Speed3
This paper proposes Rapid Diffusion, a novel framework for training and deploying super-resolution, text-to-image latent diffusion models with rich entity knowledge injected and optimized networks and employs BladeDISC and FlashAttention techniques to optimize computational graphs of the generated models for online deployment.
- 3
- 3
- ParaSum: Contrastive Paraphrasing for Low-Resource Extractive Text Summarization3
This paper proposes a novel paradigm for low-resource extractive summarization, called ParaSum, which reformulates text summarization as textual paraphrasing, and minimizes the training gap between the summarization model and PLMs, enabling a more effective probing of the knowledge encoded within PLMs and enhancing the summarizations performance.
- 3
- Refiner: Data Refining against Gradient Leakage Attacks in Federated Learning3
A novel defensive paradigm that departs from conventional gradient perturbation approaches and instead focuses on the construction of robust data is explored, and empirical evaluations demonstrate the superior defense effectiveness of Refiner at defending against state-of-the-art attacks.
- 3
- CAT-BERT: A Context-Aware Transferable BERT Model for Multi-turn Machine Reading Comprehension3
A framework to model both single-turn and multi-turn MRC tasks which allows knowledge sharing from different source MRC tasks to help solve the target MRC task is presented, and the CAT-BERT model is proposed, which jointly learns to solve both single-turn and multi-turn MRC tasks in a single pre-trained language model.
- 3
- 3
- 3
- 2
- 2
- Hallucination-Aware Prompt Optimization for Text-to-Video Synthesis2
VidPrompter is introduced, an innovative large multi-modal model supporting T2V applications with three key functionalities: generating detailed prompts from raw videos, enhancing prompts from videos grounded with short descriptions, and refining simple user-provided prompts to elevate T2V video quality.
- 2
- Learning Invariant Representations for New Product Sales Forecasting via Multi-Granularity Adversarial Learning2
A Multi-granularity AdversaRial Learning framework (MARL) is proposed to leverage knowledge from old products and improve the quality of invariant representations for more accurate sales predictions, demonstrating that the method is more effective than the existing state-of-the-art baselines for new product sales forecasting.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- TransPrompt v2: A Transferable Prompting Framework for Cross-task Text Classification–
Extensive experiments show that TransPrompt v2 outperforms single-task and cross-task strong baselines over multiple NLP tasks and datasets, and shows that the meta-learner can effectively improve the performance of PLMs on previously unseen tasks.
- –
- –
- Learning What to Ask: Mining Product Attributes for E-commerce Sales from Massive Dialogue Corpora–
AliMeMOSAIC is presented, a novel knowledge mining and conversational assistance framework that extracts core product attributes from massive dialogue corpora for better conversational recommendation experience and consistently improves the overall recommendation performance of the CRS system.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Predicting Users’ Age Range in Micro-blog Network–
This report presents the work on WISE 2013 Challenge Track II to predict the age range of Weibo users, and shows that ensemble classifiers based on AdaBoost achieves the best prediction results.
- –
- –
- –
- –
- –
- –
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 2026-10-16. 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.