Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks88 from public data
- Task-oriented Dialogue System for Automatic Diagnosis255
Experimental results on this dialogue system show that additional symptoms extracted from conversation can greatly improve the accuracy for disease identification and the dialogue system is able to collect these symptoms automatically and make a better diagnosis.
- Template-free Prompt Tuning for Few-shot NER210
This work proposes a more elegant method to reformulate NER tasks as LM problems without any templates by discarding the template construction process while maintaining the word prediction paradigm of pre-training models to predict a class-related pivot word (or label word) at the entity position.
- LoRAMoE: Alleviating World Knowledge Forgetting in Large Language Models via MoE-Style Plugin186
Experimental results show that, as the instruction data increases, Lo-RAMoE can significantly improve the ability to process downstream tasks, while maintaining the world knowledge stored in the LLM.
- Cached Long Short-Term Memory Neural Networks for Document-Level Sentiment Classification185
A Cached Long Short-Term Memory neural networks (CLSTM) to capture the overall semantic information in long texts and introduces a cache mechanism, which divides memory into several groups with different forgetting rates and thus enables the network to keep sentiment information better within a recurrent unit.
- DISC-MedLLM: Bridging General Large Language Models and Real-World Medical Consultation136
The proposed DISC-MedLLM is a comprehensive solution that leverages Large Language Models (LLMs) to provide accurate and truthful medical response in end-to-end conversational healthcare services, surpassing existing medical LLMs in both single-turn and multi-turn consultation scenarios.
- Certified Robustness to Text Adversarial Attacks by Randomized [MASK]116
This study proposes a certifiably robust defense method by randomly masking a certain proportion of the words in an input text, which can defend against not only word substitution-based attacks, but also character-level perturbations.
- Benchmark Self-Evolving: A Multi-Agent Framework for Dynamic LLM Evaluation94
A benchmark self-evolving framework to dynamically evaluate rapidly advancing Large Language Models, utilizing a multi-agent system to manipulate the context or question of original instances, reframing new evolving instances with high confidence that dynamically extend existing benchmarks.
- Training Large Language Models for Reasoning through Reverse Curriculum Reinforcement Learning88
R$^3$ progressively slides the start state of reasoning from a demonstration's end to its beginning, facilitating easier model exploration at all stages, allowing outcome supervision to offer step-level signals and precisely pinpoint errors.
- Mask Attention Networks: Rethinking and Strengthen Transformer83
A novel understanding of SAN and FFN are presented as Mask Attention Networks (MANs) and it is shown that they are two special cases of MANs with static mask matrices, which limit the capability for localness modeling in text representation learning.
- A Knowledge-Aware Sequence-to-Tree Network for Math Word Problem Solving81
Experimental results revealed that the KA-S2T model can achieve better performance than previously reported best results and use a tree-structured decoder with a state aggregation mechanism to capture the long-distance dependency and global expression information.
- 79
- SocioVerse: A World Model for Social Simulation Powered by LLM Agents and A Pool of 10 Million Real-World Users70
SocioVerse is introduced, an LLM-agent-driven world model for social simulation that can reflect large-scale population dynamics while ensuring diversity, credibility, and representativeness through standardized procedures and minimal manual adjustments.
- RMB: Comprehensively Benchmarking Reward Models in LLM Alignment62
This work proposes RMB, a comprehensive RM benchmark that covers over 49 real-world scenarios and includes both pairwise and Best-of-N (BoN) evaluations to better reflect the effectiveness of RMs in guiding alignment optimization.
- Towards Efficient NLP: A Standard Evaluation and A Strong Baseline61
The proposed ELUE has a strong Pareto Frontier and makes a better evaluation for efficient NLP models, and a strong baseline, ElasticBERT, which allows BERT to exit at any layer in both static and dynamic ways is released.
- Task-oriented Dialogue System for Automatic Disease Diagnosis via Hierarchical Reinforcement Learning.58
Experimental results on both self-constructed real-world and synthetic datasets demonstrate that the proposed hierarchical framework achieves higher accuracy in disease diagnosis compared with existing systems.
- Tasty Burgers, Soggy Fries: Probing Aspect Robustness in Aspect-Based Sentiment Analysis56
A simple but effective approach to enrich ABSA test sets is developed to disentangle the confounding sentiments of the non-target aspects from the target aspect's sentiment, and several ways to improve aspect robustness are explored.
- Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision54
This paper delves into a two-player paradigm that separates the roles of reasoning and critique models, where the critique model provides step-level feedback to supervise the reasoning (actor) model during both test-time and train-time, and proposes a critique-in-the-loop self-improvement method.
- 49
- Are Large Language Models Good Prompt Optimizers?46
This work conducts a comprehensive study to uncover the actual mechanism of LLM-based Prompt Optimization, and introduces a new "Automatic Behavior Optimization" paradigm, which directly optimizes the target model's behavior in a more controllable manner.
- 46
- ElectionSim: Massive Population Election Simulation Powered by Large Language Model Driven Agents42
ElectionSim, an innovative election simulation framework based on large language models, designed to support accurate voter simulations and customized distributions, is introduced, together with an interactive platform to dialogue with simulated voters.
- TextObfuscator: Making Pre-trained Language Model a Privacy Protector via Obfuscating Word Representations41
This paper proposes TextObfuscator, a novel framework for preserving inference privacy by applying random perturbations to clustered representations that obscures word information while retaining the original word functionality.
- Hierarchical reinforcement learning for automatic disease diagnosis32
Inspired by the offline consultation process, this work proposes to integrate a hierarchical policy structure of two levels into the dialogue system for policy learning that achieves higher accuracy and symptom recall in disease diagnosis compared with existing systems.
- Safe Inputs but Unsafe Output: Benchmarking Cross-modality Safety Alignment of Large Vision-Language Models29
The SIUO is developed, a cross-modality benchmark encompassing 9 critical safety domains, such as self-harm, illegal activities, and privacy violations, that reveals substantial safety vulnerabilities in both closed- and open-source LVLMs, underscoring the inadequacy of current models to reliably interpret and respond to complex, real-world scenarios.
- End-to-End Neural Text Classification for Tibetan29
Three end-to-end neural models for Tibetan text classification are investigated and the experimental results show that the end- to-end models outperform the traditional TibetanText classification methods.
- Transferring from Formal Newswire Domain with Hypernet for Twitter POS Tagging26
This work proposes a hypernetwork-based method to generate different parameters to separately model contexts with different expression styles for POS tagging for Twitter, and shows that this approach achieves better performance than state-of-the-art methods in most cases.
- Align Voting Behavior with Public Statements for Legislator Representation Learning24
This work constructs a heterogeneous graph for the legislative context and uses relational graph neural networks to learn the representation of legislators with the guidance of historical records of their voting and hashtag usage, and demonstrates that legislator representation learned captures nuances in statements.
- 23
- EgoNight: Towards Egocentric Vision Understanding at Night with a Challenging Benchmark21
EgoNight, the first comprehensive benchmark for nighttime egocentric vision, with visual question answering (VQA) as the core task, is presented, and EgoNight-VQA provides a strong foundation for advancing application-driven egocentric vision research and for developing models that generalize across illumination domains.
- Question Classification using Multiple Classifiers21
By using compositive statistic and rule classifiers, and by introducing dependency structure from Minipar and linguistic knowledge from Wordnet into question representation, the research shows high accuracy in question classification.
- FDU at TREC-9: CLIR, Filtering and QA Tasks.20
Fudan University takes part in the TREC-9 conference for the first time and has participated in three tracks of CLIR, Filtering and QA, where it participated in the sub-task of adaptive filtering and batch filtering.
- TableVLM: Multi-modal Pre-training for Table Structure Recognition19
A novel multi-modal pre-training model for table structure recognition, named TableVLM, which learns to capture rich table structure-related features by multiple carefully-designed unsupervised objectives inspired by the notion of masked visual-language modeling.
- FDU at TREC 2007: opinion retrieval of Blog Track18
The system consisted of the preprocess part, the topic retrieval part and sentiment analysis part, which adopted pseudo-relevance feedback and a novel approach to form a modified query.
- 15
- Constructing Phrase-level Semantic Labels to Form Multi-Grained Supervision for Image-Text Retrieval14
This paper proposes Semantic Structure Aware Multimodal Transformer (SSAMT) for multi-modal representation learning, and utilizes different kinds of attention mechanisms to enforce interactions of multi-grained semantic units in both sides of vision and language.
- MulDimIF: A Multi-Dimensional Constraint Framework for Evaluating and Improving Instruction Following in Large Language Models13
MulDimIF, a multi-dimensional constraint framework encompassing three constraint patterns, four constraint categories, and four difficulty levels, is introduced and a controllable instruction generation pipeline is designed.
- Aspect-based sentiment analysis with enhanced aspect-sensitive word embeddings13
A weakly supervised method is described to create an aspect-sensitive lexicon for each aspect, which is a relatively accurate representation of the sentiments that are related to that aspect.
- 12
- 11
- Chinese Named Entity Recognition Augmented with Lexicon Memory10
A novel fragmentbased Chinese named entity recognition (NER) model augmented with a lexicon-based memory in which both characterlevel and word-level features are combined to generate better feature representations for possible entity names is proposed.
- Automatic Term Name Generation for Gene Ontology: Task and Dataset10
This work proposes a novel task, namely term name generation for GO, and presents a graph-based generative model that incorporates the relations between genes, words and terms forterm name generation, which exhibits great advantages over the strong baselines.
- 9
- Cross-Lingual Dependency Parsing by POS-Guided Word Reordering9
This work proposes a novel approach to cross-lingual dependency parsing based on word reordering that achieves better or comparable results across 25 target languages, and outperforms a baseline by a significant margin on the languages that are greatly different from the source one.
- EliteKV: Scalable KV Cache Compression via RoPE Frequency Selection and Joint Low-Rank Projection8
EliteKV, a flexible modification framework for RoPE-based models supporting variable KV cache compression ratios, first identifies the intrinsic frequency preference of each head using RoPElite, and joint low-rank compression of key and value enables partial cache sharing.
- Negative Sample is Negative in Its Own Way: Tailoring Negative Sentences for Image-Text Retrieval8
This work argues that current retrieval-based negative sample construction approach is limited in the scale of the dataset thus fail to identify negative sample of high difficulty for every image, and proposes TAGS-DC to generate synthetic sentences automatically as negative samples.
- 8
- A Unified Model for Joint Chinese Word Segmentation and Dependency Parsing.8
Different from previous joint models, the proposed model is a graph-based model and more concise, which results in fewer efforts of feature engineering and achieves the state-of-the-art results in both Chinese word segmentation and dependency parsing.
- 8
- FDUQA on TREC 2004 QA Track.8
In this year’s QA Track, the main contribution for factoid question is two improvements for Web-based QA system and the system combination, and for definition question, effective methods are adopted to utilize both the existing definitions in the Web knowledge bases and the automatically generated structured patterns.
- 7
- 7
- Leveraging Declarative Knowledge in Text and First-Order Logic for Fine-Grained Propaganda Detection7
Experiments show that the method achieves superior performance, demonstrating that injecting declarative knowledge expressed in both natural language and first-order logic can help the model to make more accurate predictions.
- 7
- Enhancing the Capability and Robustness of Large Language Models through Reinforcement Learning-Driven Query Refinement6
A lightweight query refinement model is introduced and trained using a specially designed reinforcement learning approach that incorporates multiple objectives to enhance particular capabilities of LLMs and strengthens their robustness against jailbreak attacks.
- Modeling Evolution of Message Interaction for Rumor Resolution6
This paper model confrontation and reciprocity between message pairs via discrete variational autoencoders which effectively reflects the diversified opinion interactivity and captures the variation of message interaction using a hierarchical framework to better integrate information flow of a rumor cascade.
- Building Personalized Simulator for Interactive Search6
A simulator is proposed to employ to mimic the environment for the offline training of the agent and modelbased approach is used to train the simulator and is able to use the data efficiently.
- 6
- Question Answering for Machine Reading with Lexical Chain.6
The system follows the steps of reading comprehension as a language learner and achieves the c@1 measure of 0.28 and 0.26 for the two submissions, respectively on the QA4MRE test dataset.
- Template-independent wrapper for web forums6
Experimental results on a large number of real Web forums indicate that the proposed algorithm can correctly distinguish between the original post and replies.
- 5
- 4
- Overview of AI-Debater 2023: The Challenges of Argument Generation Tasks4
This paper organizes two tracks to handle the argumentative generation tasks in different scenarios, namely, Counter-Argument Generation and Claim-based Argument Generation, and introduces the related datasets.
- 4
- A Long Dependency Aware Deep Architecture for Joint Chinese Word Segmentation and POS Tagging.4
This work proposes a long dependency aware deep architecture for joint S&T task to simulate the feature templates of traditional discrete feature based models, and uses different filters to model the complex compositional features with convolutional and pooling layer, and utilizes long distance dependency information with recurrent layer.
- MetaRM: Shifted Distributions Alignment via Meta-Learning3
MetaRM is a method leveraging meta-learning to align the RM with the shifted environment distribution, designed to train the RM by minimizing data loss, particularly for data that can improve the differentiation ability to examples of the shifted target distribution.
- fastHan: A BERT-based Joint Many-Task Toolkit for Chinese NLP.3
This work presents fastHan, an open-source toolkit for four basic tasks in Chinese natural language processing: Chinese word segmentation, Part-of-Speech tagging, named entity recognition, and dependency parsing, which is a joint many-task model based on a pruned BERT.
- 2
- 2
- Who Responded to Whom: The Joint Effects of Latent Topics and Discourse in Conversation Structure2
A neural model is proposed to learn latent topics and discourse in word distributions, and predict pairwise initiation-response links via exploiting topic consistency and discourse dependency, which significantly outperforms the previous state of the arts.
- Overview of the CAIL 2023 Argument Mining Track2
A detailed overview of the CAIL 2023 Argument Mining Track, one of the Chinese AI and Law Challenge (CAIL) 2023 tracks, is given, which mainly uses summarized judgment documents but can also refer to trial recordings.
- Recognizing Inference in Texts with Markov Logic Networks2
This article designs specific semantic rules based on the surface, syntactic, and semantic representations of texts, and map these rules to logical representations, and utilizes Markov logic networks framework to make predictions with combining statistical and logical reasoning.
- 1
- Effective Length Extrapolation via Dimension-Wise Positional Embeddings Manipulation1
Dimension-Wise Positional Embeddings Manipulation (DPE), a training-free framework to extrapolate the context window of LLMs by diving into RoPE's different hidden dimensions, significantly surpasses well-known baselines such as YaRN and Self-Extend.
- Detecting Essence Code Clones via Information Theoretic Analysis1
ECScan is introduced, a novel detection tool that leverages information theory to assess the semantic importance of code lines and significantly outperforms existing tools in detecting essence clones, achieving an average F1-score of 85%.
- 1
- 1
- 1
- ProofInfer: Generating Proof via Iterative Hierarchical Inference1
The proposed ProofInfer proposes a divide-and-conquer algorithm to encode the proof tree as the plain text without losing structure information, and achieves comparable performance to the state-of-the-art model with about 40% of the training data.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
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 2026-10-14. 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.