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
- Min ZhangSuggested from co-authorship
- Qun LiuSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works230 from public data
- Variational Neural Machine Translation211
This paper builds a neural posterior approximator conditioned on both the source and the target sides, and equip it with a reparameterization technique to estimate the variational lower bound, and shows that the proposed variational neural machine translation achieves significant improvements over the vanilla neural machinetranslation baselines.
- A Novel Graph-based Multi-modal Fusion Encoder for Neural Machine Translation177
A novel graph-based multi-modal fusion encoder for NMT that captures various semantic relationships between multi- modal semantic units (words and visual objects) and provides an attention-based context vector for the decoder.
- Cascade Grouped Attention Network for Referring Expression Segmentation160
A Cascade Grouped Attention Network with two innovative designs: Cascade Grouping Attention (CGA) and Instance-level Attention (ILA) loss that can achieve the high efficiency of one-stage RES while possessing a strong reasoning ability comparable to the two-stage methods.
- 131
- 129
- Accelerating Neural Transformer via an Average Attention Network129
The proposed average attention network is applied on the decoder part of the neural Transformer to replace the original target-side self-attention model and enables the neuralTransformer to decode sentences over four times faster than its original version with almost no loss in training time and translation performance.
- Asynchronous Bidirectional Decoding for Neural Machine Translation125
This paper equips the conventional attentional encoder-decoder NMT framework with a backward decoder, in order to explore bidirectional decoding for NMT, and achieves substantial improvements over the conventional NMT by 3.14 and 1.38 BLEU points, respectively.
- Shallow Convolutional Neural Network for Implicit Discourse Relation Recognition112
A Shallow Convolutional Neural Network (SCNN) is proposed, which contains only one hidden layer but is effective in relation recognition, which alleviates the overfitting problem, while the convolution and nonlinear operations help preserve the recognition and generalization ability of the model.
- Progressive Self-Supervised Attention Learning for Aspect-Level Sentiment Analysis92
A progressive self-supervised attention learning approach for neural ASC models, which automatically mines useful attention supervision information from a training corpus to refine attention mechanisms, and augment the conventional training objective with a regularization term.
- 92
- A Real-Time Global Inference Network for One-Stage Referring Expression Comprehension86
The proposed RealGIN outperforms most existing methods and achieves very competitive performances against the most advanced one, i.e., MAttNet, under the same hardware, and can boost the processing speed by 10–20 times over the existing methods.
- 86
- 85
- Variational Recurrent Neural Machine Translation84
A novel variational recurrent neural machine translation model that introduces a series of latent random variables to model the translation procedure of a sentence in a generative way, instead of a single latent variable.
- Multi-Domain Neural Machine Translation with Word-Level Domain Context Discrimination81
This paper jointly model NMT with monolingual attention-based domain classification tasks and improves NMT as follows, distinguishing and exploiting word-level domain contexts for multi-domain NMT.
- 79
- Exploiting Temporal Relationships in Video Moment Localization with Natural Language78
A novel Temporal Compositional Modular Network (TCMN) where a tree attention network first automatically decomposes a sentence into three descriptions with respect to the main event, context event and temporal signal is proposed.
- Graph-based Neural Sentence Ordering71
A novel and flexible graph-based neural sentence ordering model, which adopts graph recurrent network \cite{Zhang:acl18} to accurately learn semantic representations of the sentences, which outperforms the existing state-of-the-art systems on several benchmark datasets.
- 70
- 69
- Enhanced aspect-based sentiment analysis models with progressive self-supervised attention learning66
This work proposes a progressive self-supervised attention learning approach for attentional ABSA models that enhances the performance of all three models and integrates the proposed approach into three state-of-the-art neuralABSA models.
- Structural Information Preserving for Graph-to-Text Generation59
This work introduces two types of autoencoding losses, each individually focusing on different aspects (a.k.a. views) of input graphs, that can guide the model for preserving input information via multi-task training.
- When to use Graphs in RAG: A Comprehensive Analysis for Graph Retrieval-Augmented Generation57
This work proposes GraphRAG-Bench, a comprehensive benchmark designed to evaluate GraphRAG models on both hierarchical knowledge retrieval and deep contextual reasoning, and systematically investigates the conditions when GraphRAG surpasses traditional RAG.
- A Robust Adversarial Training Approach to Machine Reading Comprehension55
This paper proposes a novel robust adversarial training approach to improve the robustness of MRC models in a more generic way and boosts the overall performance on adversarial datasets and outperforms the state-of-the-art methods.
- Translation Model Adaptation for Statistical Machine Translation with Monolingual Topic Information53
This paper proposes a novel approach for translation model adaptation by utilizing in-domain monolingual topic information instead of the in- domain bilingual corpora, which incorporates the topic information into translation probability estimation.
- Neural Collective Entity Linking Based on Recurrent Random Walk Network Learning47
A novel end-to-end neural network with recurrent random-walk layers for collective EL, which introduces external knowledge to model the semantic interdependence between different EL decisions and achieves better performance than other state-of-the-art models.
- BACO: A Background Knowledge- and Content-Based Framework for Citing Sentence Generation45
BACO is presented, a BAckground knowledge- and COntent-based framework for citing sentence generation, which considers two types of information: background knowledge by leveraging structural information from a citation network; and content, which represents in-depth information about what to cite and why to cite.
- A Structure Self-Aware Model for Discourse Parsing on Multi-Party Dialogues43
A Structure Self-Aware (SSA) model is proposed, which adopts a novel edge-centric Graph Neural Network to update the information between each Elementary Discourse Unit (EDU) pair layer by layer, so that expressive representations can be learned without historical predictions.
- Multi-modal neural machine translation with deep semantic interactions43
This model extends the conventional multi-modal NMT by introducing the following two attention neural networks: a bi-directional attention network for modeling text and image representations, where the semantic representations of text are learned by referring to the image representation, and vice versa.
- Dynamic Capsule Attention for Visual Question Answering43
Inspired by the recent work of Capsule Network, CapsAtt treats visual features as capsules and obtains the attention output via dynamic routing, which updates the attention weights by calculating coupling coefficients between the underlying and output capsules to make the model much more compact.
- Bilingual attention based neural machine translation41
A novel bilingual attention based NMT, where its bilingual attention mechanism exploits decoding history and enables the NMT model to better dynamically select and exploit source side and target side information.
- Leveraging Dependency Forest for Neural Medical Relation Extraction41
A graph neural network is used to represent the forests, automatically distinguishing the useful syntactic information from parsing noise, and shows that this method outperforms the standard tree-based methods, giving the state-of-the-art results in the literature.
- Lattice-to-sequence attentional Neural Machine Translation models41
Lattice-to-sequence attentional NMT models are proposed, which generalize the standard Recurrent Neural Network encoders to lattice topology and take as input a word lattice which compactly encodes many tokenization alternatives, and learn to generate the hidden state for the current step from multiple inputs and hidden states in previous steps.
- StructCap41
The proposed StructCap model parses a given image into key entities and their relations organized in a visual parsing tree, which is transformed and embedded under an encoder-decoder framework via visual attention.
- Modeling Discourse Structure for Document-level Neural Machine Translation39
A HAN encoder based on a hierarchical attention network is proposed to improve document-level NMT with the aid of discourse structure information and can significantly outperform both Transformer and Transformer+HAN on the English-to-German dataset.
- A Label Dependence-Aware Sequence Generation Model for Multi-Level Implicit Discourse Relation Recognition38
This paper considers multi-level IDRR as a conditional label sequence generation task and proposes a Label Dependence-aware Sequence Generation Model (LDSGM) for it, which designs a label attentive encoder to learn the global representation of an input instance and its level-specific contexts.
- Exploring Contextual Word-level Style Relevance for Unsupervised Style Transfer37
A novel attentional sequence-to-sequence (Seq2seq) model that dynamically exploits the relevance of each output word to the target style for unsupervised style transfer and fine-tune this model using a carefully-designed objective function involving style transfer, style relevance consistency, content preservation and fluency modeling loss terms.
- Exploiting reverse target-side contexts for neural machine translation via asynchronous bidirectional decoding37
This paper extends the conventional attentional encoder-decoder NMT framework by introducing a backward decoder, in order to explore asynchronous bidirectional decoding for NMT, and achieves significant improvements over the conventional RNNSearch.
- LiteSearch: Efficacious Tree Search for LLM36
This study introduces a novel guided tree search algorithm with dynamic node selection and node-level exploration budget (maximum number of children) calculation to tackle the issue of wasteful search strategies in LLM.
- 36
- Iterative Dual Domain Adaptation for Neural Machine Translation36
An iterative dual domain adaptation framework for NMT that first pretrain in-domain and out-of-domain NMT models using their own training corpora respectively, and iteratively perform bidirectional translation knowledge transfer based on knowledge distillation until the in- domain NMT model convergences.
- 35
- 34
- Fog-Based Two-Phase Event Monitoring and Data Gathering in Vehicular Sensor Networks33
An efficient continuous event-monitoring and data-gathering framework based on fog nodes in vehicular sensor networks that adaptively adjusts the threshold to upload a suitable amount of data for decision making, while at the same time suppressing unnecessary message transmissions is proposed.
- Enhancing Pointer Network for Sentence Ordering with Pairwise Ordering Predictions32
This work proposes to enhance the pointer network decoder by using two pairwise ordering prediction modules: the FUTURE module predicts the relative orientations of other unordered sentences with respect to the candidate sentence, and the HISTORY module measures the local coherence between several sentences without the influence of noisy left-side context.
- Improving Tree-Structured Decoder Training for Code Generation via Mutual Learning31
This paper first throughly analyze the context modeling difference between neural code generation models with different traversals based decodings (preorder traversal vs breadth-first traversal), and then proposes to introduce a mutual learning framework to jointly train these models.
- From statistical methods to deep learning, automatic keyphrase prediction: A survey30
This work comprehensively summarize representative studies from the perspectives of dominant models, datasets and evaluation metrics, focusing highly on deep learning-based keyphrase prediction, which attracts increasing attention of this task in recent years.
- A neural generative autoencoder for bilingual word embeddings30
A neural generative bilingual autoencoder (NGBAE) which introduces a latent variable to explicitly induce the underlying semantics of bilingual text to obtain better BWEs from more robust bilingual semantics by modeling the semantic distributions ofilingual text.
- 28
- Neural Simile Recognition with Cyclic Multitask Learning and Local Attention28
A novel cyclic multitask learning framework for neural simile recognition, which stacks the subtasks and makes them into a loop by connecting the last to the first.
- 27
- K-armed Bandit based Multi-Modal Network Architecture Search for Visual Question Answering26
The experimental results show that with the help of KAB-NAS, AGAN can achieve the state-of-the-art performance on both benchmark datasets with much fewer parameters and computations.
- A neural image captioning model with caption-to-images semantic constructor26
This paper extends the conventional CNN-LSTM image captioning model by introducing a caption-to-images semantic reconstructor, which reconstructs the semantic representations of the input image and its similar images from hidden states of the decoder.
- EpiCoder: Encompassing Diversity and Complexity in Code Generation25
A novel feature tree-based synthesis framework, which revolves around hierarchical code features derived from high-level abstractions of code, that provides precise control over the complexity of the generated code, enabling functionalities that range from function-level operations to multi-file scenarios.
- Plenty is Plague: Fine-Grained Learning for Visual Question Answering25
This article proposes a fine-grained VQA learning paradigm with an actor-critic based learning agent, termed FG-A1C, which includes a learning agent that adaptively and intelligently schedules the most difficult question types in each training epoch.
- Bilingually-constrained Synthetic Data for Implicit Discourse Relation Recognition24
This work proposes to use bilingually-constrained synthetic implicit data from a bilingual sentence-aligned corpus according to the implicit/explicit mismatch be-tween different languages for implicit discourse relation recognition.
- 23
- Exploring Dynamic Selection of Branch Expansion Orders for Code Generation23
This paper proposes to equip the Seq2Tree model with a context-based Branch Selector, which is able to dynamically determine optimal expansion orders of branches for multi-branch nodes, and optimize the selector through reinforcement learning, and formulate the reward function as the difference of model losses obtained through different expansion orders.
- 23
- Bilingual Correspondence Recursive Autoencoder for Statistical Machine Translation23
This paper proposes a new neural network model called Bilingual Correspondence Recursive Autoencoder (BCorrRAE) to model bilingual phrases in translation and incorporates word alignments into BCorrRAE to allow it freely access bilingual constraints at different levels.
- Bilingual Correspondence Recursive Autoencoders for Statistical Machine Translation23
This paper proposes a new neural network model called Bilingual Correspondence Recursive Autoencoder (BCorrRAE) to model bilingual phrases in translation and incorporates word alignments into BCorrRAE to allow it freely access bilingual constraints at different levels.
- 22
- 22
- Bridging Subword Gaps in Pretrain-Finetune Paradigm for Natural Language Generation22
A plug-and-play embedding generator is introduced to produce the representation of any input token, according to pre-trained embeddings of its morphologically similar ones, leading to more efficient and better performed downstream NLG models.
- Towards Making the Most of Dialogue Characteristics for Neural Chat Translation22
This paper proposes to promote the chat translation by introducing the modeling of dialogue characteristics into the NCT model, and designs four auxiliary tasks including monolingual response generation, cross-lingual responsegeneration, next utterance discrimination, and speaker identification.
- Towards User-Driven Neural Machine Translation22
Experimental results confirm that the proposed user-driven NMT can generate user-specific translations and contribute the first Chinese-English parallel corpus annotated with user behavior called UDT-Corpus.
- Hierarchical multi-task learning with CRF for implicit discourse relation recognition22
This paper proposes a hierarchical multi-task neural network with a conditional random field layer (HierMTN-CRF) for multi-level IDRR that achieves significantly better and more consistent results over several competitive baselines on multi- level IDRR, without additional time overhead.
- Variational Neural Discourse Relation Recognizer22
This paper introduces neural discourse relation models to approximate the prior and posterior distributions of the latent variable, and employs these approximated distributions to optimize a reparameterized variational lower bound that allows VarNDRR to be trained with standard stochastic gradient methods.
- Who To Align With: Feedback-Oriented Multi-Modal Alignment in Recommendation Systems21
This study suggests that multi-modal alignment in MRSs should be in a one-way direction, and presents a plug-and-play framework, called FETTLE, which significantly improves various backbone models.
- Confidence Based Bidirectional Global Context Aware Training Framework for Neural Machine Translation20
A Confidence Based Bidirectional Global Context Aware (CBBGCA) training framework for NMT, where the NMT model is jointly trained with an auxiliary conditional masked language model (CMLM) and pertinently incorporate bidirectional global context to the N MT model on its unconfidently-predicted target words via knowledge distillation.
- BattRAE: Bidimensional Attention-Based Recursive Autoencoders for Learning Bilingual Phrase Embeddings20
A bidimensional attention based recursive autoencoder (BattRAE) is proposed to integrate clues and sourargetinteractions at multiple levels of granularity into bilingualphrase representations to evaluate the effectiveness of this state-of-the-art SMT system.
- On the Cultural Gap in Text-to-Image Generation19
A novel multi-modal metric that considers object-text alignment to filter the fine-tuning data in the target culture, which is used to fine-tune a T2I model to improve cross-cultural generation is proposed.
- CASA: Conversational Aspect Sentiment Analysis for Dialogue Understanding19
The task of conversational aspect sentiment analysis (CASA) is introduced that can provide useful fine-grained sentiment information for dialogue understanding and planning and extends the standard aspect-based sentiment analysis to the conversational scenario with several major adaptations.
- 18
- 18
- 18
- Improving Implicit Discourse Relation Recognition with Discourse-specific Word Embeddings18
A simple and effective method to learn discourse-specific word embeddings (DSWE) for implicit discourse relation recognition and on the PDTB data set, using DSWE as features achieves significant improvements over baselines.
- Retrieval-Augmented Multi-Modal Chain-of-Thoughts Reasoning for Large Language Models17
This paper introduces a novel approach that addresses this challenge by using retrieval mechanisms to dynamically select demonstration examples based on cross-modal and intra-modal similarities and is the first to apply Retrieval-Augmented Generation with CoT to complex multi-modal reasoning tasks.
- Don't Get Lost in the Trees: Streamlining LLM Reasoning by Overcoming Tree Search Exploration Pitfalls17
FETCH is a flexible, plug-and-play system compatible with various tree search algorithms, which mitigates over-exploration by merging semantically similar states using agglomerative clustering of text embeddings obtained from a fine-tuned SimCSE model and tackles under-exploration by incorporating temporal difference learning with adjusted $\lambda$-returns during training to reduce variance.
- Controllable Dialogue Generation With Disentangled Multi-Grained Style Specification and Attribute Consistency Reward17
This paper defines and categorizes the commonly-used control attributes into global and local ones, which possess different granularities of effects on response generation, and proposes a controllable dialogue generation model to steer response generation under multi-attribute constraints.
- 17
- POS Tag-enhanced Coarse-to-fine Attention for Neural Machine Translation17
An NMT model with tag-enhanced attention mechanism, where NMT and POS tagging are jointly modeled via multi-task learning and word prediction is facilitated by simultaneously utilizing the context vector from fine attention and the predicted target POS tags.
- Learning better discourse representation for implicit discourse relation recognition via attention networks17
A full attention model that combines their strengths into a unified framework is proposed that significantly benefits from highlighting relation-relevant words and yields competitive and even better results against several state-of-the-art systems.
- A Context-Aware Topic Model for Statistical Machine Translation17
A context-aware topic model is proposed for lexical selection that not only models local contexts and global topics but also captures their correlations, and a Gibbs sampling algorithm is introduced for statistical estimation and inference.
- Exploring Self-Distillation Based Relational Reasoning Training for Document-Level Relation Extraction16
A document-level RE model with a reasoning module that contains a core unit, the reasoning multi-head self-attention unit, which can cover more relational triples than previous methods and which consistently outperforms all competitive baselines.
- Simplifying Neural Machine Translation with Addition-Subtraction Twin-Gated Recurrent Networks16
Experiments on WMT14 translation tasks demonstrate that ATR-based neural machine translation can yield competitive performance on English-German and English-French language pairs in terms of both translation quality and speed.
- Search-engine-augmented dialogue response generation with cheaply supervised query production15
A dialogue model that can access the vast and dynamic information from any search engine for response generation, and a training algorithm using cheap noisy supervision for the query producer, where the signals are obtained by comparing retrieved articles with the next dialogue response.
- An External Knowledge Enhanced Graph-based Neural Network for Sentence Ordering14
A novel and flexible external knowledge enhanced graph-based neural network for sentence ordering, where various kinds of relations are exploited to make the graph representation more expressive and less noisy.
- A Survey on Multi-modal Machine Translation: Tasks, Methods and Challenges13
An exhaustive overview of 99 prior works is offered, comprehensively summarizing representative studies from the perspectives of dominant models, datasets, and evaluation metrics to provide researchers with a better understanding of multi-modal machine translation's current state.
- 13
- Continual Learning for Multilingual Neural Machine Translation via Dual Importance-based Model Division13
A dual importance-based model division method to divide the model parameters into two parts and separately model the translation of the original and new tasks and demonstrates that this method consistently outperforms several strong baselines under different incremental translation scenarios.
- A Multi-Task Multi-Stage Transitional Training Framework for Neural Chat Translation13
A multi-task multi-stage transitional (MMT) training framework, where an NCT model is trained using the bilingual chat translation dataset and additional monolingual dialogues, and introduces the modelling of dialogue coherence and speaker characteristic into the NCT model.
- An Iterative Multi-Source Mutual Knowledge Transfer Framework for Machine Reading Comprehension13
This paper proposes a novel iterative multi-source mutual knowledge transfer framework for MRC that can continuously enhance all domain-specific MRC models by enabling each model to iteratively and differentially absorb the domain-shared knowledge from others.
- More Than An Answer13
This paper proposes a novel network architecture, termed Neural Pivot Network (NPN), towards simultaneous VQA and generating explanations in a multi-task learning architecture that performs significantly better than alternatives and state-of-the-art schemes trained on V QA datasets only.
- Domain Adaptation for Conversational Query Production with the RAG Model Feedback12
This work improves the algorithm by taking the advance of retrieval-augmented generation (RAG) and exploring several practical techniques such as knowledge distillation for stable training and performs significantly better especially in a more challenging setting over strong baselines.
- 12
- 12
- Learning Lexicalized Reordering Models from Reordering Graphs12
This work proposes to use a structure named reordering graph, which represents all phrase segmentations of a sentence pair, to learn lexicalized reordering models efficiently, and shows that this approach significantly outperforms the baseline method.
- Revolutionizing Database Q&A with Large Language Models: Comprehensive Benchmark and Evaluation11
DQABench is introduced, the first comprehensive database QA benchmark for LLMs and reveals the strengths and limitations of nine open-source and commercial LLMs, and the impact of various service components (e.g., fine-tuning, QCR, RAG, TIG).
- EmoTrans: Emotional Transition-based Model for Emotion Recognition in Conversation11
An Emotional Transition-based Emotion Recognizer (EmoTrans) is proposed and new state-of-the-art results obtained on MELD and IEMOCAP, demonstrating the superiority of EmoTrans.
- HyperNetwork-based Decoupling to Improve Model Generalization for Few-Shot Relation Extraction11
This paper proposes a HyperNetwork-based Decoupling approach to improve the generalization of FSRE models and designs a two-step training strategy along with a class-agnostic aligner, by which the generated classifiers focus on acquiring relation-specific knowledge and the encoder is encouraged to learn more general relation knowledge.
- 11
- Social Media Based Topic Modeling for Smart Campus: A Deep Topical Correlation Analysis Method11
A novel deep topical correlation analysis (DTCA) approach is presented, which achieves robust and accurate topic detection for microblogs and simultaneously handles the two challenges aforementioned.
- Recurrent Neural Machine Translation11
A novel recurrent neural machine translation (RNMT), which not only preserves the ability to model varying-length source annotations but also better captures long-term dependencies, and can implicitly reflect the alignment to source sentence.
- AVG-LLaVA: An Efficient Large Multimodal Model with Adaptive Visual Granularity10
This work introduces AVG-LLaVA, an LMM that can adaptively select the appropriate visual granularity based on the input image and instruction, and proposes RGLF, a novel training paradigm that aims at aligning the granularity predicted by the router with the preferences of the LMM, without the need for additional manually annotated data.
- 10
- Otem&Utem: Over- and Under-Translation Evaluation Metric for NMT10
Two quantitative metrics are proposed, the Otem and Utem, to automatically evaluate the system performance in terms of over- and under-translation respectively, based on the proportion of mismatched n-grams between gold reference and system translation.
- Not All Languages are Equal: Insights into Multilingual Retrieval-Augmented Generation9
Experimental results reveal linguistic inequalities and offer advice for improving multilingual Retrieval Augmented Generation and cross-lingual knowledge transfer as well as underscore the complexities inherent in multilingual RALMs.
- Fine-Grained Self-Endorsement Improves Factuality and Reasoning9
This work proposes a self-endorsement framework that leverages the fine-grained fact-level comparisons across multiple sampled responses that can better alleviate hallucinations, especially for longform generation tasks.
- 9
- KGR4: Retrieval, Retrospect, Refine and Rethink for Commonsense Generation9
A novel Knowledge-enhanced Commonsense Generation framework, termed KGR4, consisting of four stages: Retrieval, Retrospect, Refine, Rethink, which selects the output sentence from candidate sentences produced by generators with different hyper-parameters.
- Improving Graph-based Sentence Ordering with Iteratively Predicted Pairwise Orderings9
A novel sentence ordering framework which introduces two classifiers to make better use of pairwise orderings for graph-based sentence ordering (Yin et al. 2019, 2021), and achieves state-of-the-art performance.
- 9
- A Topic‐Triggered Translation Model for Statistical Machine Translation9
This paper extends the conventional translation model to a topic-triggered one, and estimates topic-specific translation probabilities of translation rules by leveraging topical context information, and online score selected translation rules according to topic posterior distributions of translated sentences.
- LagNet: Deep Lagrangian Mechanics for Plug-and-Play Molecular Representation Learning8
Experimental results show that LagNet can well learn 3D molecular structure features, and outperforms previous state-of-the-art baselines related molecular representation by a significant margin.
- CLLE: A Benchmark for Continual Language Learning Evaluation in Multilingual Machine Translation8
This work proposes the first Continual Language Learning Evaluation benchmark CLLE in multilingual translation and proposes a novel framework COMETA based on Constrained Optimization and META-learning to alleviate catastrophic forgetting and dependency on historical training data by using a meta-model to retain the important parameters for old languages.
- Neural Machine Translation with Decoding History Enhanced Attention8
A DHEA mechanism is reformulated to render NMT model better at selecting both source-side and target-side information, and achieves comparable results with the state-of-the-art.
- LegalGraphRAG: Multi-Agent Graph Retrieval-Augmented Generation for Reliable Legal Reasoning7
The approach introduces two core components: a hierarchical legal graph that hierarchically organizes legal sources to enable retrieval at appropriate abstraction levels, and a multi-agent system for reliable legal reasoning, where a Researcher retrieves candidate evidence, an Auditor rigorously verifies its validity against source documents, and an Adjudicator synthesizes the set of verified evidence to render a final judgment.
- 7
- FactGen: Faithful Text Generation by Factuality-aware Pre-training and Contrastive Ranking Fine-tuning7
A factuality-aware pretraining-finetuning framework named FactGen, which fully considers factuality during two training stages, and utilizes a natural language inference model to construct target texts that are entailed by the source texts, resulting in a more factually consistent pre-training objective.
- Exploring All-In-One Knowledge Distillation Framework for Neural Machine Translation7
This work proposes a novel All-In-One Knowledge Distillation (AIO-KD) framework for NMT, which generates multiple satisfactory students at once and designs a two-stage mutual learning strategy, which alleviates the negative impacts of poorly-performed students on the early-stage student interactions.
- 7
- Alignment-consistent recursive neural networks for bilingual phrase embeddings7
This paper considers word alignments as a semantic bridge between the source and target phrases, and proposes two neural networks based on the conventional recursive autocoder, which exploitword alignments to generate alignment-consistent bilingual phrase structures.
- Graph-Based Collective Lexical Selection for Statistical Machine Translation7
A graph-based framework for collective lexical selection is established on a translation graph that captures not only local associations between source-side content words and their target translations but also targetside global dependencies in terms of relatedness among target items.
- LiteSearch: Efficient Tree Search with Dynamic Exploration Budget for Math Reasoning6
This study introduces a novel guided tree search algorithm with a goal-directed heuristic function and node-level exploration budget and node-level exploration budget (maximum number of children) calculation to tackle the issue of waste in tree search.
- Locate-and-Focus: Enhancing Terminology Translation in Speech Language Models6
Experimental results demonstrate that the proposed novel Locate-and-Focus method effectively locates terminologies within utterances and enhances the success rate of terminology translation, while maintaining robust general translation performance.
- Exploring Optimal Transport-Based Multi-Grained Alignments for Text-Molecule Retrieval6
This work introduces the Optimal Transport-based Multi-grained Alignments model (ORMA), a novel approach that facilitates multi-grained alignments between textual descriptions and molecules, and is the first attempt to explore alignments at both the motif and multi-token levels.
- Signer Diversity-driven Data Augmentation for Signer-Independent Sign Language Translation6
The Signer Diversity-driven Data Augmentation (SDDA) method is proposed that can achieve good generalization without relying on signer identity labels, and introduces a consistency loss and a discrimination loss to enhance the learning of signer-independent features.
- Modeling different effects of user and product attributes on review sentiment classification6
A simple and effective review sentiment classification model based on hierarchical attention network, where different effects of user and product attributes are respectively captured via fusion modules and attention mechanisms is proposed.
- 6
- Towards Better Multi-modal Keyphrase Generation via Visual Entity Enhancement and Multi-granularity Image Noise Filtering6
A novel multi-modal keyphrase generation model is proposed, which not only enriches the model input with external knowledge, but also effectively filters image noise.
- Bridging the Gap between Training and Inference: Multi-Candidate Optimization for Diverse Neural Machine Translation6
This paper proposes a multi-candidate optimization framework to score the diversity and the quality of candidate translations during training, and optimize the diverse NMT model with two strate-gies based on reinforcement learning, namely hard constrained training and soft constrained training.
- 6
- Enhancing Chinese Word Segmentation via Pseudo Labels for Practicability6
Experiments show that the proposed semi-supervised neural method can not only keep the practi-cability of the lightweight student model but also improve the performance of segmentation effectively and is reliable and practical as a pre-processing step of the downstream NLP tasks at the minimum cost.
- Uncovering Media Bias via Social Network Learning6
This article realizes the first attempt of predicting the latent bias of media outlets by analyzing their social network structures by adopting a graph-based label propagation scheme to supplement the missing links and augment label information for model training.
- Boosting implicit discourse relation recognition with connective-based word embeddings6
This paper learns connective-based word embeddings (CBWE) by performing connective classification on explicit discourse data and achieves the state-of-the-art performance on both the English PDTB and Chinese CDTB data sets.
- Convolution-Enhanced Bilingual Recursive Neural Network for Bilingual Semantic Modeling.6
A convolution-enhanced bilingual recursive neural network (ConvBRNN) is proposed, which not only exploits word alignments to guide the generation of phrase structures but also integrates multiple-level information of the generated phrase structures into bilingual semantic modeling.
- Unsupervised word sense induction using rival penalized competitive learning6
This paper constructs multi-granularity semantic spaces to learn the representations of ambiguous instances, in order to capture richer semantic knowledge during context modeling to improve the quality of WSI over several competitive baselines.
- Regularized Structured Perceptron: A Case Study on Chinese Word Segmentation, POS Tagging and Parsing6
Three simple but effective task-independent regularization methods are studied: to average weights of different trained models to reduce the bias caused by the specific order of the training examples, to add penalty term to the loss function and to randomly corrupt the data flow during training.
- 6
- 5
- UniGenCoder: Merging SEQ2SEQ and SEQ2TREE Paradigms for Unified Code Generation5
This paper proposes UniGenCoder for code-related generation tasks, which consists of a shared encoder, a shared decoder with a minimal set of additional parameters to unify two paradigms, and a selector that dynamically chooses optimal paradigm for each instance.
- 5
- TempParaphraser: “Heating Up” Text to Evade AI-Text Detection through Paraphrasing5
Temp-Paraphraser is proposed, a simple yet effective paraphrasing framework that simulates high-temperature sampling effects through multiple normal-temperature generations, effectively evading detection.
- Towards better text image machine translation with multimodal codebook and multi-stage training5
This work proposes a multimodal codebook based TIMT model that is equipped with a multimodal codebook that effectively associates images with relevant texts, thus providing useful supplementary information for translation.
- Hard-Synth: Synthesizing Diverse Hard Samples for ASR using Zero-Shot TTS and LLM5
This paper proposes Hard-Synth, a novel ASR data augmentation method that leverages large language models (LLMs) and advanced zero-shot TTS to clone speech styles that the ASR model finds challenging to recognize.
- 5
- Multi-Level Cross-Modal Alignment for Speech Relation Extraction5
A Multi-level Cross-modal Alignment Model (MCAM) for SpeechRE, which can effectively train the alignment adapter to bridge the modality gap between the speech encoder and the text decoder and demonstrates the efficacy of the method.
- Exploring Effective Inter-Encoder Semantic Interaction for Document-Level Relation Extraction5
A document-level RE model with a Graph-Transformer Network (GTN), which includes two core sublayers: the graph-attention sublayer that simultaneously models global and local contexts of nodes in the document graph and the structure-aware adversarial knowledge distillation, by which it can effectively transfer the structural information of GTN to the document encoder.
- Adaptive Token-level Cross-lingual Feature Mixing for Multilingual Neural Machine Translation5
This paper proposes a novel token-level feature mixing method that enables the model to capture different features and dynamically determine the feature sharing across languages, and can be extended to zero-shot translation.
- 5
- Free VQA Models from Knowledge Inertia by Pairwise Inconformity Learning5
A Pairwise Inconformity Learning (PIL) is proposed to tackle the issue of knowledge inertia in visual question answering (VQA) and strengthens the importance of visual features in prediction with a novel dynamic-margin based triplet loss that efficiently increases the semantic discrepancies between pos./neg.
- Co-training for Implicit Discourse Relation Recognition Based on Manual and Distributed Features5
A co-training approach based on manual features and distributed features, which identifies useful instances from these artificial data to enlarge the labeled data and significantly outperforms the baselines.
- 5
- A Dual-Perspective Metaphor Detection Framework Using Large Language Models4
DMD is proposed, a novel dual-perspective framework that harnesses both implicit and explicit applications of metaphor theories to guide LLMs in metaphor detection and adopts a self-judgment mechanism to validate the responses from the aforementioned forms of guidance.
- LLM-OREF: An Open Relation Extraction Framework Based on Large Language Models4
An OpenRE framework based on large language models (LLMs) is proposed, which directly predicts new relations for test instances by leveraging their strong language understanding and generation abilities, without human intervention.
- 4
- 4
- 4
- Mitigating the negative impact of over-association for conversational query production4
This work carefully analyzes the negative effects of the over-association phenomenon on pretrained Seq2seq query producers and proposes effective instance-level weighting strategies for training to mitigate these issues from multiple perspectives and shows that this model selects better concepts from dialogue histories and is 10 times more data efficient than the baseline.
- 4
- OpenFact: Factuality Enhanced Open Knowledge Extraction4
This work focuses on the factuality property during the extraction of an OpenIE corpus named OpenFact, which contains more than 12 million high-quality knowledge triplets and proposes a comprehensive framework to handle both expressiveness and groundedness.
- 4
- 4
- AAN+: Generalized Average Attention Network for Accelerating Neural Transformer4
A generalized average attention network (AAN+) aiming at speeding up decoding by reducing the dependency from O(m) to O(1) is proposed, finding that the learned self-attention weights in the decoder follow some patterns which can be approximated via a dynamic structure.
- 4
- Cross-lingual implicit discourse relation recognition with co-training4
A cross-lingual implicit DRR framework that exploits an available English corpus for the Chinese DRR task, using machine translation to generate Chinese instances from a labeled English discourse corpus and training two classifiers in Chinese and English in a co-training way.
- 4
- A Topic-Triggered Language Model for Statistical Machine Translation4
A novel topic-triggered language model is described, which takes into account the topical context by estimating the n-gram probability under the given topics and online adjusts language model score according to topic distributions.
- Chinese Named Entity Recognition and Disambiguation Based on Wikipedia4
The experiments show that the use of Wikipedia features can improve the accuracy of named entity recognition and achieve named entity disambiguation using Wikipedia disambigsuation pages and contextual information.
- PLaST: Towards Paralinguistic-aware Speech Translation3
This work proposes Paralinguistic-aware Speech Translation (PLaST), a novel dual-branch framework which directly leverages paralinguistic cues beyond the linguistic content and outperforms the strong baseline with an average of 5.5 global contrastive likelihood scores on the paralinguistic-sensitive benchmark ContraProST.
- DetectRL-X: Towards Reliable Multilingual and Real-World LLM-Generated Text Detection3
DetectRL-X is introduced, a comprehensive multilingual benchmark designed to evaluate advanced detectors across 8 dimensions and analyzed how domains, generators, attack strategies, text length, and refinement operations influence performance in different languages, underscoring DetectRL-X as an effective benchmark for strengthening multilingual and language-specific detectors.
- Beyond Black-Box Interventions: Latent Probing for Faithful Retrieval-Augmented Generation3
ProbeRAG is proposed, a novel framework for faithful RAG that operates in three stages: fine-grained knowledge pruning to filter irrelevant context, latent conflict probing to identify hard conflicts in the model's latent space, and conflict-aware attention to modulate attention heads toward faithful context integration.
- 3
- 3
- Inhibition of joint inflammation in rheumatoid arthritis by Wuwei Ganlu Medicinal Bath Granules via blocking toll-like receptor 4/nuclear factor-kappa B and NOD-like Receptor Pyrin Domain-Containing 3 signaling3
This study aimed to identify the active constituents of WGMG as well as understand the pharmacological mechanisms of WGMG in the treatment of collagen-induced arthritis (CIA) in rats, and found that 68 overlapping targets between WGMG components and RA were revealed.
- 3
- 3
- 3
- Code-Aware Cross-Program Transfer Hyperparameter Optimization3
CaTHPO is proposed, a code-aware cross-program transfer hyperparameter optimization framework, which makes three improvements, and presents a safe controller to dynamically prune undesirable sample points based on tuning experiences of similar programs.
- Bridging the Domain Gaps in Context Representations for k-Nearest Neighbor Neural Machine Translation3
This work designs a reviser to revise the key representations of kNN-MT, making them better fit for the downstream domain, and demonstrates that this method can effectively boost the datastore retrieval and translation quality of knn-MT.
- 3
- 3
- Context-Aware Phrase Representation for Statistical Machine Translation3
This work proposes a context-aware phrase representation learning framework, which extends the bilingually-constrained recursive autoencoder with context modeling component and introduces the bipartite network embedding method to learn better word and topic embedding, which further improve the quality of the phrase representation.
- Leveraging bilingually-constrained synthetic data via multi-task neural networks for implicit discourse relation recognition3
This paper proposes to use bilingually-constrained synthetic implicit data (BiSynData) to enrich the training data, which can alleviate the drawbacks of SynData and designs a multi-task neural network model to incorporate the authors' BiSynData to benefit implicit discourse relation recognition.
- Discriminative reordering model adaptation via structural learning3
This paper proposes a novel adaptive discriminative reordering model (DRM) based on structural learning, which can capture correspondences among reordering features from two different domains, and learns a shared feature representation for cross-domain phrase reordering.
- Lightweight Transformer for sEMG Gesture Recognition with Feature Distilled Variational Information Bottleneck2
This paper proposes the feature distilled variational information bottleneck (FDVIB), a method that leverages knowledge distillation to learn from a high-precision teacher model at levels of feature and prediction, significantly reducing parameters and computations, and simplifying the structure.
- 2
- 2
- 2
- A Lexical Resource-Constrained Topic Model for Word Relatedness2
The proposed lexical resource-constrained topic model is an extension of probabilistic latent semantic analysis, which automatically learns word-level distributed representations forward relatedness measurement and introduces generalized expectation maximization (GEM) algorithm for statistical estimation.
- Quantum Digital Signature Scheme Based on Finite Automata2
This paper proposes a new quantum digital signature scheme based on finite automata and shows its security against existing attacks.
- Bilingual Autoencoders with Global Descriptors for Modeling Parallel Sentences2
This paper explores a bilingual autoencoder approach to model parallel sentences by extracting sentence-level global descriptors from word embeddings, and constructing two monolingual autoencoders over these descriptors on the source and target language.
- 2
- Dependency-Based Bracketing Transduction Grammar for Statistical Machine Translation2
A novel dependency-based bracketing transduction grammar for statistical machine translation, which converts a source sentence into a target dependency tree, and employs two different maximum entropy models to determine the reordering and combination of partial dependency structures.
- MemGraphRAG: Memory-based Multi-Agent System for Graph Retrieval-Augmented Generation1
MemGraphRAG is a novel framework that introduces a memory-based multi-agent system to ensure high-quality graph construction and outperforms the state-of-the-art baseline models with comparable efficiency.
- 1
- 1
- m 3 BERT: A Modern, Multi-lingual, Matryoshka Bidirectional Encoder1
m3BERT: a Modern, Multi-lingual, Matryoshka Bidirectional Encoder, which features a novel pretraining strategy that jointly optimizes representations across both transformer layers and multiple embedding dimensions enables a single model to be tailored to varied resource and accuracy targets while maintaining consistency with pretraining.
- 1
- EditEval: Towards Comprehensive and Automatic Evaluation for Text-guided Video Editing1
This work proposes EditEval, the largest-ever video editing benchmark to comprehensively evaluate the performance of video editing models in three aspects: Textual Faithfulness, Frame Consistency, and Video Fidelity, and proposes EditScore, leveraging the advanced reasoning and comprehension capabilities of Multi-modal Large Language Models as evaluators to assess edited videos across the aforementioned aspects.
- 1
- On the Information Redundancy in Non-Autoregressive Translation1
This work revisits the multi-modal problem in recently proposed NAT models and identifies two types of information redundancy errors that correspond well to lexical and reordering multi-modality problems.
- 1
- 1
- 1
- Learning Spatiotemporal and Geometric Features with ISA for Video-Based Facial Expression Recognition1
Experiments on CK+ and MMI facial expression databases show that recognition performance can be improved effectively by incorporating geometric features into spatiotemporal features and the overall accuracy of this method is comparable to some deep learning based methods.
- Topic-aware pivot language approach for statisticalmachine translation1
This study proposes two topic-aware pivot language approaches to use different levels of pivot-side context, and builds an interpolated model bringing the above methods together to further enhance the system performance.
- Pivot Probability Induction for Statistical Machine Translation with Topic Similarity1
The latent topic structure of the document-level training data is learned automatically and each phrase translation rule is assigned to a topic distribution, allowing the translation system to consider the semantic relatedness among different rules.
- Improving Chinese Word Segmentation Using Partially Annotated Sentences1
A learning task that can use partially annotated data as the training data is described that improves the performances of both supervised model and semi-supervised models.
- –
- –
- –
- GIFT: Guided Fine-Tuning and Transfer for Enhancing Instruction-Tuned Language Models–
GIFT (Guided Fine-Tuning and Transfer), a simple and efficient framework that incorporates guidance from the instruction model into task adaptation, consistently outperforms direct fine-tuning and representative transfer-based baselines, while maintaining robust generalization and favorable test-time scaling behavior.
- –
- –
- –
- –
- BioPro: Towards Difference-Aware Gender Fairness for Vision-Language Models–
This work proposes BioPro (Bias Orthogonal Projection), an entirely training-free framework that effectively reduces gender bias in neutral cases while maintaining gender faithfulness in explicit ones, providing a promising direction toward achieving selective fairness in VLMs.
- Gloss Matters: Unlocking the Potential of Non-Autoregressive Sign Language Translation–
To alleviate the inconsistency between training and inference of GLevT, which is introduced by glosses, a dual-centric learning policy and a keyframe-based gloss replacement method for training are proposed, further improving the translation quality of GLevT.
- –
- –
- –
- –
- Semantically Smooth Bilingual Phrase Embeddings Based on Recursive Autoencoders–
This work extends the conventional bilingual recursive autoencoders by preserving the translation and paraphrase probability distributions via regularization terms to simultaneously exploit richer explicit and implicit similarity constraints for bilingual phrase embeddings.
- –
- Topic-Guided Automatical Human-Simulated Tweeting System–
In this paper, a human-simulated tweeting system is first designed to generate the multiple and appropriate tweets for given topics and the effectiveness of the proposed system and the superiority of the specific image captioning model are evaluated by sufficient quantitative comparisons and qualitative analysis in a real-world Twitter dataset.
- Tree-state based rule selection models for hierarchical phrase-based machine translation–
This paper proposes tree-state models to discriminate the good or bad usage of translation rules based on the syntactic structures of the source sentence and proposes to use statistical models and context dependent features to estimate the probability of each tree state for each translation rule.
- –
- Graph-based lexicalized reordering models for statistical machine translation–
This paper proposes a method to take the number of adjacent phrases into account for better estimation of reordering models, and uses a compact structure named reordering graph to represent all phrase segmentations of a parallel sentence.
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.