EMNLP 2023 Papers — Page 4
Conference on Empirical Methods in Natural Language Processing · 1047 papers
Document-Level Machine Translation with Large Language Models
Longyue Wang (Tencent AI Lab), Zhaopeng Tu (Tencent AI Lab)
GenerationTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This paper systematically evaluates the performance of large language models (ChatGPT GPT-3.5 and GPT-4) in document-level machine translation, conducting in-depth experiments from three aspects: prompt engineering, model comparison, and discourse modeling.
Document-level Relationship Extraction by Bidirectional Constraints of Beta Rules
Yichun Liu (Tianjin University), Yaxin Li (Tianjin University)
Graph Neural NetworkTransformerLarge Language ModelText
🎯 What it does: Propose a framework named BCBR that enhances document-level relation extraction models by utilizing a Beta rule miner and bidirectional logical constraints;
Does the Correctness of Factual Knowledge Matter for Factual Knowledge-Enhanced Pre-trained Language Models?
Boxi Cao (Chinese Information Processing Laboratory), Le Sun (Chinese Information Processing Laboratory)
Explainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningTextGraph
🎯 What it does: This paper introduces a counterfactual knowledge perturbation analysis framework to systematically examine the causal impact of injecting factual knowledge into pre-trained language models (PLMs) on the performance of downstream tasks.
Don’t Take This Out of Context!: On the Need for Contextual Models and Evaluations for Stylistic Rewriting
Akhila Yerukola (Carnegie Mellon University), Maarten Sap (Carnegie Mellon University)
GenerationTransformerLarge Language ModelPrompt EngineeringTextSequential
🎯 What it does: Studied the necessity of introducing context in stylized text rewriting, proposed a contextual rewriting model, a context-based evaluation method, and introduced a new evaluation metric called CtxSimFit.
Don’t Trust ChatGPT when your Question is not in English: A Study of Multilingual Abilities and Types of LLMs
Xiang Zhang (University of Alberta), Grzegorz Kondrak (University of Alberta)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Systematically evaluate the multilingual capabilities of large language models using both quantitative and qualitative methods, and experimentally test GPT's performance on translation equivalence and variation tasks through prompt translation and response back-translation methods.
Doolittle: Benchmarks and Corpora for Academic Writing Formalization
Shizhe Diao (Hong Kong University of Science and Technology), Tong Zhang (Hong Kong University of Science and Technology)
GenerationTransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Propose the academic writing formalization (AWF) task and the non-parallel dataset DOOLITTLE, using multi-objective reinforcement learning to enhance text quality.
DPP-TTS: Diversifying prosodic features of speech via determinantal point processes
Seongho Joo (Seoul National University), Kyomin Jung (Seoul National University)
GenerationTransformerFlow-based ModelAudio
🎯 What it does: Proposed a text-to-speech model called DPP‑TTS based on Determinantal Point Processes (DPP), enhancing speech expressiveness diversity through diversified intonation and rhythm.
Dr ChatGPT tell me what I want to hear: How different prompts impact health answer correctness
Bevan Koopman (CSIRO), Guido Zuccon (University of Queensland)
Explainability and InterpretabilityTransformerPrompt EngineeringTextBiomedical DataRetrieval-Augmented Generation
🎯 What it does: This study evaluates the accuracy of ChatGPT in answering health information questions and explores the impact of different prompting methods (question-only prompting vs. prompting with retrieved results) on the correctness of answers.
DREAM: Deployment of Recombination and Ensembles in Argument Mining
Florian Ruosch (University of Zurich), Abraham Bernstein (University of Zurich)
ClassificationRecognitionTextBenchmark
🎯 What it does: Proposed and implemented the DREAM framework, which combines existing Argument Mining systems to enhance the accuracy of various tasks
DSI++: Updating Transformer Memory with New Documents
Sanket Vaibhav Mehta (Carnegie Mellon University), Donald Metzler (Google Research)
RetrievalTransformerText
🎯 What it does: Proposes DSI++, a method for continual learning and incremental indexing within the Differentiable Search Indices (DSI) framework, addressing the problem of requiring model retraining when new documents are added.
Dual-Channel Span for Aspect Sentiment Triplet Extraction
Pan Li (Southwest Petroleum University), Kai Zhang (East China Normal University)
Graph Neural NetworkTransformerText
🎯 What it does: This paper proposes the Dual-Span model, which employs dual-channel span generation (syntactic relations and part-of-speech relations) and dual graph attention networks for end-to-end extraction of sentiment triplets (aspect terms, opinion terms, sentiment polarity).
Dual-Feedback Knowledge Retrieval for Task-Oriented Dialogue Systems
Tianyuan Shi (Sun Yat-sen University), Qifan Wang (Sun Yat-sen University)
RetrievalTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose a bidirectional feedback retrieval framework that separates retrieval and generation in task-oriented dialogue systems, using positive and negative feedback generated by the generator to train the retriever.
DueT: Image-Text Contrastive Transfer Learning with Dual-adapter Tuning
Taku Hasegawa (NTT Human Informatics Laboratories), Kuniko Saito (NTT Human Informatics Laboratories)
ClassificationRetrievalTransformerContrastive LearningMultimodality
🎯 What it does: Inserting a gated adapter unit (GAU) into frozen pre-trained unimodal encoders (ViT, BERT), and only training these adapters to complete image-text contrastive learning for transfer learning.
DUMB: A Benchmark for Smart Evaluation of Dutch Models
Wietse de Vries (University of Groningen), Malvina Nissim (University of Groningen)
Hyperparameter SearchTransformerSupervised Fine-TuningTextBenchmark
🎯 What it does: Created the DUMB (Dutch Model Benchmark) benchmark, which includes 9 Dutch NLP tasks, and introduced the Relative Error Reduction (RER) evaluation metric.
DUnE: Dataset for Unified Editing
Afra Akyürek (Boston University), Derry Wijaya (Boston University)
Data-Centric LearningLarge Language ModelSupervised Fine-TuningTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Construct a unified editing dataset DUNE that covers four types of editing scenarios: factual, reasoning, arithmetic, and bias.
Dynamic Top-k Estimation Consolidates Disagreement between Feature Attribution Methods
Jonathan Kamp (Vrije Universiteit Amsterdam), Antske Fokkens (Vrije Universiteit Amsterdam)
ClassificationExplainability and InterpretabilityTransformerText
🎯 What it does: Propose a dynamic method to determine the k value for selecting important terms in the explanations of text classifiers, avoiding bias and errors caused by fixed k.
Dynosaur: A Dynamic Growth Paradigm for Instruction-Tuning Data Curation
Da Yin (Ucla), Kai-Wei Chang (Ucla)
Data-Centric LearningTransformerLarge Language ModelText
🎯 What it does: Proposes the DYNOSAUR framework, which leverages metadata from Huggingface datasets and large language models (LLMs) to automatically generate instruction-based training data.
E-CORE: Emotion Correlation Enhanced Empathetic Dialogue Generation
Fengyi Fu (University of Science and Technology of China), Zhendong Mao (University of Science and Technology of China)
GenerationTransformerTextBenchmark
🎯 What it does: Propose the E-CORE framework, which leverages multi-resolution sentiment maps and a sentiment-related enhanced decoder to learn sentiment correlations through explicit modeling and supervision, thereby improving empathetic dialogue generation;
e-THERAPIST: I suggest you to cultivate a mindset of positivity and nurture uplifting thoughts
Kshitij Mishra (Indian Institute of Technology Patna), Asif Ekbal (Indian Institute of Technology Patna)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Developed a psychological therapy dialogue system called e-THERAPIST capable of generating polite and friendly conversations based on user gender, age, personality traits, and emotions.
EasyQuant: An Efficient Data-free Quantization Algorithm for LLMs
Hanlin Tang, Zhanhui Kang
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Proposes EasyQuant, a data-free and training-free LLM weight-only quantization method.
EDeR: Towards Understanding Dependency Relations Between Events
Ruiqi Li (Australian National University), Leyang Cui (Tencent AI lab)
ClassificationRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper constructs a human-annotated event dependency dataset called EDeR, trains a prediction model on this dataset, and subsequently demonstrates performance improvements on semantic role labeling and coreference resolution tasks.
EDIS: Entity-Driven Image Search over Multimodal Web Content
Siqi Liu (Cornell University), William Wang (University of California Santa Barbara)
RetrievalTransformerSupervised Fine-TuningVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: Propose a large-scale (millions of scale) news domain entity-driven image search dataset called EDIS and study how to retrieve multimodal candidates (image + title).
Editing Common Sense in Transformers
Anshita Gupta (University of Massachusetts Amherst), Niket Tandon (Allen Institute for AI)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed MEMITCSK, an improved parameter editing algorithm for directly correcting common-sense misjudgments in Transformer models, and constructed a new evaluation set called PROBE SET;
Editing Large Language Models: Problems, Methods, and Opportunities
Yunzhi Yao (Zhejiang University), Ningyu Zhang (Zhejiang University)
TransformerLarge Language ModelPrompt EngineeringTextReview/Survey PaperBenchmark
🎯 What it does: Systematic review and comparison of large language model (LLM) editing methods, proposing more comprehensive evaluation metrics (portability, locality, and efficiency) and constructing the corresponding benchmark dataset.
Effects of sub-word segmentation on performance of transformer language models
Jue Hou (University of Helsinki), Roman Yangarber (University of Helsinki)
TransformerLarge Language ModelText
🎯 What it does: This paper compares subword segmentation algorithms such as BPE, Morfessor, and StateMorph, investigating their impact on Transformer language models (GPT, BERT) across multiple languages (Finnish, Russian, English, Turkish), including perplexity, convergence speed, downstream task performance, and model scale efficiency.
Efficient Algorithms for Recognizing Weighted Tree-Adjoining Languages
Alexandra Butoi (ETH Zürich), David Chiang (University of Notre Dame)
RecognitionComputational Efficiency
🎯 What it does: This paper proposes a new algorithm for two-level formalisms of Tree-Adjoining languages (e.g., CFG/CFG, PDA/CFG), which can efficiently compute the total weight of strings and all derivations (stringsum and allsum).
Efficient Classification of Long Documents via State-Space Models
Peng Lu (Huawei Noah's Ark Lab), Ivan Kobyzev (Huawei Noah's Ark Lab)
ClassificationText
🎯 What it does: For the long document classification task, we propose using a state space model (SSM) to achieve efficient classification.
Efficient Grammatical Error Correction Via Multi-Task Training and Optimized Training Schedule
Andrey Bout (Huawei Noah's Ark Lab), Irina Piontkovskaya (Huawei Noah's Ark Lab)
GenerationData SynthesisOptimizationComputational EfficiencyTransformerSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: This paper proposes leveraging alignment information between original and corrected sentences by decomposing the GEC task into several sequence-to-sequence subtasks for joint learning through multi-task pre-training and optimized training scheduling; further improving model performance by designing the training data order (no shuffling, arranging datasets sequentially, keeping sentences from the same article in the same batch).
Elaborative Simplification as Implicit Questions Under Discussion
Yating Wu (University of Texas at Austin), Junyi Jessy Li (University of Texas at Austin)
GenerationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: This paper studies elaborative simplification in text simplification, treating elaborations as answers to implicit questions (QUD), constructing the ELABQUD dataset, and conducting experiments in a two-step process (question generation → elaboration generation) using GPT-3 for zero-shot elaboration generation.
Elevating Code-mixed Text Handling through Auditory Information of Words
Mamta (Indian Institute of Technology Patna), Asif Ekbal (Indian Institute of Technology Patna)
ClassificationExplainability and InterpretabilityAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose two pre-training methods, SMLM (SOUNDEX Masked Language Modeling) and SAMLM (SOUNDEX Aligned Masked Language Modeling), integrating SOUNDEX phoneme encoding with word vectors into BERT and RoBERTa to enhance robustness and classification performance on code-mixed text.
Emergence of Abstract State Representations in Embodied Sequence Modeling
Tian Yun (Brown University), Chen Sun (Brown University)
Representation LearningRobotic IntelligenceTransformerReinforcement LearningTextSequential
🎯 What it does: Investigated whether abstract state representations can emerge in embodied sequence modeling, validated using BabyAI tasks.
Empathy Intent Drives Empathy Detection
Liting Jiang (Xinjiang University), Wushour Slamu (Xinjiang University)
ClassificationRecognitionTransformerLarge Language ModelText
🎯 What it does: This paper introduces a co-sympathetic intent recognition task and jointly trains empathy detection, significantly improving the accuracy of empathy detection.
Empirical Study of Zero-Shot NER with ChatGPT
Tingyu Xie (Zhejiang University), Hongwei Wang (Zhejiang University)
RecognitionTransformerPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper conducts systematic experiments on ChatGPT's performance in zero-shot named entity recognition (NER) tasks and proposes four reasoning-based technical strategies: Task Decomposition (Decomposed-QA), Syntactic Enhancement (Grammatical Prompting and Tool Augmentation), Two-Stage Self-Consistency Voting (SC), and their combinations.
Empower Nested Boolean Logic via Self-Supervised Curriculum Learning
Hongqiu Wu (Shanghai Jiao Tong University), Min Zhang (Soochow University)
TransformerLarge Language ModelTextBenchmarkChain-of-Thought
🎯 What it does: Designed a self-supervised curriculum learning method called CLR to train language models to master multi-layer nested Boolean logic reasoning, and proposed the corresponding BoolKill benchmark dataset.
Enabling Large Language Models to Generate Text with Citations
Tianyu Gao (Princeton University), Danqi Chen (Princeton University)
GenerationRetrievalTransformerLarge Language ModelPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Designed and implemented the ALCE benchmark, constructing an end-to-end system enabling large language models (LLMs) to automatically retrieve evidence, generate answers with citations, and provide three-dimensional automatic evaluation (fluency, correctness, citation quality).
End-to-End Single-Channel Speaker-Turn Aware Conversational Speech Translation
Juan Pablo Zuluaga-Gomez (Idiap Research Institute), Marcello Federico (AWS AI Labs)
TransformerTextAudio
🎯 What it does: This paper proposes an end-to-end multi-speaker multi-turn dialogue speech translation system (STAC-ST), which simultaneously accomplishes ASR, speech translation, and speaker switch detection through multi-task training, achieving joint output of tasks using special marker sequences;
End-to-end Task-oriented Dialogue: A Survey of Tasks, Methods, and Future Directions
Libo Qin (School of Computer Science and Engineering, Central South University), Min Li (Research Center for Social Computing and Information Retrieval)
Large Language ModelReinforcement LearningPrompt EngineeringTextMultimodalityReview/Survey PaperBenchmarkRetrieval-Augmented Generation
🎯 What it does: This paper reviews the development of end-to-end task-oriented dialogue (EToD) systems, proposes a unified perspective and new classification (modular end-to-end vs. fully end-to-end), and organizes public resources and experimental benchmarks.
Enhancing Biomedical Lay Summarisation with External Knowledge Graphs
Tomas Goldsack (University of Sheffield), Chenghua Lin (University of Manchester)
GenerationGraph Neural NetworkTransformerBiomedical DataRetrieval-Augmented Generation
🎯 What it does: By constructing article-specific knowledge graphs for each biomedical article and injecting them into a long-text generation model, the readability and explainability of automated lay abstracts were enhanced.
Enhancing Chat Language Models by Scaling High-quality Instructional Conversations
Ning Ding (Tsinghua University), Bowen Zhou (Tsinghua University)
Data-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark
🎯 What it does: Constructed a large-scale, human-query-free multi-turn dialogue dataset named UltraChat with 1.5 million dialogues, and fine-tuned LLaMA-13B on this dataset to generate UltraLM;
Enhancing Code-Switching for Cross-lingual SLU: A Unified View of Semantic and Grammatical Coherence
Zhihong Zhu (Peking University), Yuexian Zou (Peking University)
Domain AdaptationRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose a framework named SOGO for improving semantic and syntactic consistency in zero-shot cross-lingual SLU tasks through enhanced code-switching.
Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization
Janghwan Lee (Hanyang University), Jungwook Choi (Hanyang University)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: For post-training quantization of LLMs, AQAS, SLAC, and dINT are proposed to maintain almost full performance under 4-bit weights and 8-bit activations.
Enhancing Generative Retrieval with Reinforcement Learning from Relevance Feedback
Yujia Zhou, Ji-Rong Wen (Renmin University of China)
RetrievalReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Propose a generative retrieval framework GenRRL based on reinforcement learning, which transforms the document generation task into a reinforcement learning problem and uses a relevance reward model to provide feedback on the generated docid, thereby improving retrieval ranking quality.
Enhancing Low-resource Fine-grained Named Entity Recognition by Leveraging Coarse-grained Datasets
Su Lee, Woohwan Jung (Hanyang University)
RecognitionData-Centric LearningTransformerSupervised Fine-TuningText
🎯 What it does: For low-resource fine-grained named entity recognition, a Fine-to-Coarse (F2C) mapping matrix is constructed using coarse-grained data, combined with inconsistency filtering to achieve joint training;
Enhancing Structured Evidence Extraction for Fact Verification
Zirui Wu (Peking University), Yansong Feng (Peking University)
RetrievalGraph Neural NetworkTransformerTabularBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: In open-domain fact verification, the SEE-ST method is proposed, improving table extraction and cell selection to enhance the efficiency of structured evidence retrieval.
Enhancing Task-oriented Dialogue Systems with Generative Post-processing Networks
Atsumoto Ohashi (Nagoya University), Ryuichiro Higashinaka (Nagoya University)
GenerationReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Proposed a generative post-processing network (GenPPN) that post-processes outputs from the natural language generation (NLG) module in task-oriented dialogue systems, optimized through reinforcement learning with dialogue action contribution (DA contribution) rewards.
Enhancing Textbooks with Visuals from the Web for Improved Learning
Janvijay Singh (Georgia Institute of Technology), Mrinmaya Sachan (ETH Zürich)
RetrievalVision Language ModelImageText
🎯 What it does: Leverage vision-language models (e.g., CLIP) to automatically retrieve and assign web images to textbook chapters, enhancing the visual appeal of educational materials.
Enhancing the Ranking Context of Dense Retrieval through Reciprocal Nearest Neighbors
George Zerveas (Brown University), Carsten Eickhoff (University of Tübingen)
RetrievalText
🎯 What it does: Propose an evidence label smoothing method based on reciprocal nearest neighbors (rNN) to improve the training of dense retrieval models under sparse annotations, and perform efficient re-ranking using rNN during the inference phase.
Enhancing Uncertainty-Based Hallucination Detection with Stronger Focus
Tianhang Zhang (Shanghai Jiaotong University), Luoyi Fu (Shanghai Jiaotong University)
Anomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose a reference-free, uncertainty-based hallucination detection method that simulates the human inspection process of focusing on key terms, preposition propagation, and token attribute correction;
EntSUMv2: Dataset, Models and Evaluation for More Abstractive Entity-Centric Summarization
Dhruv Mehra (Bloomberg), Daniel Preotiuc-Pietro (Bloomberg)
GenerationData-Centric LearningTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Investigated abstract entity-centric summaries, released the shorter and more abstract ENTSUMV2 dataset, and evaluated multiple models
EpiK-Eval: Evaluation for Language Models as Epistemic Models
Gabriele Prato (Mila), Sarath Chandar (Mila)
TransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposes the EpiK-Eval benchmark, specifically designed to evaluate the ability of large language models in knowledge integration (knowledge consensus), particularly merging information from segmented narratives.
Establishing Trustworthiness: Rethinking Tasks and Model Evaluation
Robert Litschko (LMU Munich), Barbara Plank (LMU Munich)
Explainability and InterpretabilityKnowledge DistillationLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This is a position paper that rethinks NLP tasks and model evaluation in the context of the growing popularity of large language models (LLMs), proposing a 'credibility' framework that emphasizes enhancing understanding of model capabilities and origins through 'knowledge facets'.
EtiCor: Corpus for Analyzing LLMs for Etiquettes
Ashutosh Dwivedi (Indian Institute of Technology Kanpur), Ashutosh Modi (Indian Institute of Technology Kanpur)
Large Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Constructed a 36k English etiquette corpus named EtiCor covering five regions: East Asia, India, the Middle East and Africa, North America and Europe, and Latin America, and proposed an 'etiquette sensitivity' evaluation task;
Euphemistic Abuse – A New Dataset and Classification Experiments for Implicitly Abusive Language
Michael Wiegand (Alpen-Adria University), Josef Ruppenhofer (FernUniversity)
ClassificationTransformerLarge Language ModelText
🎯 What it does: Construct and utilize a novel euphemistic abuse dataset aimed at identifying implicit attack statements targeting non-identity groups.
Evaluating and Modeling Attribution for Cross-Lingual Question Answering
Benjamin Muller (INRIA Paris), Xinyi Wang (Google Research)
RetrievalExplainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: Address the answer attribution problem in cross-lingual question answering systems, constructing the XOR-AttriQA dataset and evaluating the attribution level of existing cross-lingual open-retrieval QA models (CORA);
Evaluating Bias and Fairness in Gender-Neutral Pretrained Vision-and-Language Models
Laura Cabello (University of Copenhagen), Desmond Elliott (University of Copenhagen)
TransformerSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: Investigate gender bias amplification and fairness in Vision-and-Language (V&L) models, quantify bias in pre-training and fine-tuning stages, and explore mitigating bias through additional epochs of gender-neutral pre-training.
Evaluating Cross-Domain Text-to-SQL Models and Benchmarks
Mohammadreza Pourreza (University of Alberta), Davood Rafiei (University of Alberta)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextTabularBenchmark
🎯 What it does: Evaluate and rewrite cross-domain Text-to-SQL benchmarks, re-evaluate model performance, and reveal the limitations of evaluation metrics and benchmarks.
Evaluating Evaluation Metrics: A Framework for Analyzing NLG Evaluation Metrics using Measurement Theory
Ziang Xiao (Microsoft Research Montréal), Q. Vera Liao (Microsoft Research Montréal)
TextBenchmark
🎯 What it does: Propose the METRICEVAL framework, which evaluates the reliability and validity of NLG assessment metrics using measurement theory.
Evaluating Large Language Models on Controlled Generation Tasks
Jiao Sun (University of Southern California), Xuezhe Ma (University of Southern California)
GenerationLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Systematically evaluated the controllability of large language models on ten controlled generation tasks and proposed a new numerical planning benchmark.
Evaluating Object Hallucination in Large Vision-Language Models
Yifan Li (Renmin University of China), Ji-Rong Wen (Renmin University of China)
Large Language ModelVision Language ModelMultimodalityBenchmark
🎯 What it does: Systematically evaluate the target object hallucination problem in large vision-language models (LVLM), finding that their hallucination rate is higher than traditional models and investigating the reasons;
Evaluating the Rationale Understanding of Critical Reasoning in Logical Reading Comprehension
Akira Kawabata (Asahi Shimbun Company), Saku Sugawara (National Institute of Informatics)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark
🎯 What it does: Constructed a benchmark dataset named RULE, containing main questions and corresponding auxiliary sub-questions extracted from ReClor, to evaluate models' understanding of implicit reasoning in critical thinking.
Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks
Andrea Sottana (NetMind.AI), Zheng Yuan (King's College London)
GenerationTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Comprehensively evaluate the performance of large language models on three types of sequence-to-sequence tasks (text summarization, simplification, and grammar error correction), combining traditional reference-based evaluation, human assessment, and GPT-4 model evaluation, revealing significant deviations between automatic metrics and human assessments;
Evaluation of African American Language Bias in Natural Language Generation
Nicholas Deas (Columbia University), Kathleen McKeown (Columbia University)
GenerationExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Evaluate the understanding and generation biases of large language models (LLMs) toward African American Language (AAL) using two tasks: corresponding generation and masked span prediction;
Event Causality Extraction via Implicit Cause-Effect Interactions
Jintao Liu (Chinese Academy of Sciences), Xiaoyu Li (University of Chinese Academy of Sciences)
GenerationKnowledge DistillationTransformerPrompt EngineeringTextFinance Related
🎯 What it does: This paper proposes the ICE framework, modeling the event causality extraction task as a templated conditional generation problem, and capturing implicit causal interactions through teacher-student knowledge distillation and event-level optimal transport;
Event Ontology Completion with Hierarchical Structure Evolution Networks
Pengfei Cao (Chinese Academy of Sciences), Jun Zhao (China Merchants Bank)
TransformerLarge Language ModelContrastive LearningText
🎯 What it does: Propose the Event Ontology Completion (EOC) task and design the HALTON model to achieve event clustering, hierarchical expansion, and type naming
Event-Location Tracking in Narratives: A Case Study on Holocaust Testimonies
Eitan Wagner (Hebrew University of Jerusalem), Omri Abend (Hebrew University of Jerusalem)
TransformerLarge Language ModelSupervised Fine-TuningTextSequential
🎯 What it does: Proposed the event location tracking task and designed, implemented, and evaluated multiple document-level models on approximately 1000 oral histories of Jewish Holocaust survivors.
Exchange-of-Thought: Enhancing Large Language Model Capabilities through Cross-Model Communication
Zhangyue Yin (Fudan University), Xipeng Qiu (Fudan University)
Computational EfficiencyTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Proposes the Exchange-of-Thought (EoT) framework, enabling large language models (LLMs) to enhance reasoning capabilities by sharing reasoning chains through cross-model communication.
Expand, Highlight, Generate: RL-driven Document Generation for Passage Reranking
Arian Askari (Leiden University), Suzan Verberne (Leiden University)
RetrievalTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextBenchmark
🎯 What it does: Propose two pipelines, DocGen and DocGen-RL, for generating synthetic documents from queries to enhance unannotated information retrieval training data;
EXPLAIN, EDIT, GENERATE: Rationale-Sensitive Counterfactual Data Augmentation for Multi-hop Fact Verification
Yingjie Zhu (Southeast University), Yulan He (King's College London)
GenerationData SynthesisExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Generate counterfactual data for multi-hop fact verification tasks to help models avoid over-reliance on input surface features, thereby enhancing robustness to out-of-distribution data.
Explaining Interactions Between Text Spans
Sagnik Ray Choudhury (University of Michigan), Isabelle Augenstein (University of Copenhagen)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Create and use the multi-annotator SpanEx dataset, manually annotate cross-sentence span interactions in natural language inference (NLI) and fact checking (FC) tasks, and evaluate the decision-making process of large language models (LLMs) based on this dataset; propose an unsupervised method based on community detection to automatically extract interpretable text span interactions.
Explaining with Contrastive Phrasal Highlighting: A Case Study in Assisting Humans to Detect Translation Differences
Eleftheria Briakou (Google), Marine Carpuat (University of Maryland)
Explainability and InterpretabilityTransformerContrastive LearningText
🎯 What it does: Proposed a post-hoc interpretable method that uses contrastive phrase highlighting to explain predictions of NLP models comparing two texts, focusing on identifying translation differences;
Explanation Selection Using Unlabeled Data for Chain-of-Thought Prompting
Xi Ye (University of Texas at Austin), Greg Durrett (University of Texas at Austin)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Propose a black-box search framework based on unlabeled data to automatically generate and optimize chain-of-thought (explanation) prompts, thereby improving the performance of large language models on text reasoning tasks.
Explicit Planning Helps Language Models in Logical Reasoning
Hongyu Zhao (University of Chicago), Hongyuan Mei (Toyota Technological Institute at Chicago)
TransformerLarge Language ModelContrastive LearningText
🎯 What it does: Proposed the LEAP system, which utilizes language models for multi-step logical reasoning and incorporates explicit planning during the reasoning process.
Exploiting Asymmetry for Synthetic Training Data Generation: SynthIE and the Case of Information Extraction
Martin Josifoski (EPFL), Robert West (EPFL)
Data SynthesisTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: This paper proposes a process for generating synthetic training data through reverse tasks, utilizing a large language model (LLM) to first sample structured outputs (entity–relation–object triplets) and then prompt the LLM to generate corresponding natural language text, thereby obtaining high-quality, balanced closed information extraction (cIE) training samples.
Explore-Instruct: Enhancing Domain-Specific Instruction Coverage through Active Exploration
Fanqi Wan (Sun Yat Sen University), Shuming Shi (Tencent Ai Lab)
Data-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Construct domain-specific instruction sets and perform instruction tuning by combining active exploration (lookahead, backtracking) with LLMs to enhance model performance in specific domains.
Exploring All-In-One Knowledge Distillation Framework for Neural Machine Translation
Zhongjian Miao (Xiamen University), Min Zhang (Soochow University)
Knowledge DistillationTransformerText
🎯 What it does: Propose an All-In-One Knowledge Distillation framework named AIO-KD, which can generate multiple student models simultaneously while optimizing both the teacher and students at the same time;
Exploring Chain of Thought Style Prompting for Text-to-SQL
Chang-Yu Tai, Huan Sun (Ohio State University)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextTabularBenchmarkChain-of-Thought
🎯 What it does: Systematically investigate the effectiveness of Chain-of-Thought (CoT) prompting in text-to-SQL parsing, and propose the QDecomp+InterCOL prompting method based on problem decomposition;
Exploring Discourse Structure in Document-level Machine Translation
Xinyu Hu (Wangxuan Institute of Computer Technology, Peking University), Xiaojun Wan (Wangxuan Institute of Computer Technology, Peking University)
GenerationRepresentation LearningTransformerTextSequential
🎯 What it does: Explore the impact of discourse structure on translation quality in document-level machine translation, and propose a multi-granularity attention mechanism based on RST.
Exploring Distributional Shifts in Large Language Models for Code Analysis
Shushan Arakelyan (University of Southern California), Xiang Ren (University of Southern California)
Domain AdaptationMeta LearningAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: Investigate the generalization ability of large language models (CodeT5, Codex, ChatGPT) on code summarization and generation tasks under distribution shifts across different domains such as organizations, projects, and modules, and propose a retrieval-based adaptation method without labeled data to improve performance in low-data scenarios.
Exploring Jiu-Jitsu Argumentation for Writing Peer Review Rebuttals
Sukannya Purkayastha (Technical University of Darmstadt), Iryna Gurevych (Technical University of Darmstadt)
ClassificationGenerationDomain AdaptationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Constructed the JITSUPEER dataset and proposed a rebuttal generation task for peer review based on the 'Jiu-Jitsu' argumentation framework, focusing on attitude origins and theme-driven rebuttals; designed three subtasks (feasibility scoring, description generation, terminal generation) and conducted baseline experiments.
Exploring Linguistic Probes for Morphological Generalization
Jordan Kodner (Stony Brook University), Sarah Payne (Stony Brook University)
Explainability and InterpretabilityRepresentation LearningRecurrent Neural NetworkTransformerTextBenchmark
🎯 What it does: Designed and evaluated language-specific morphological probes to examine the morphological generalization capabilities of different models on English, Spanish, and Swahili.
Exploring the Boundaries of GPT-4 in Radiology
Qianchu Liu (Microsoft Health Futures), Javier Alvarez-Valle (Microsoft Health Futures)
TransformerLarge Language ModelPrompt EngineeringBiomedical DataBenchmarkChain-of-Thought
🎯 What it does: Construct a radiology text task evaluation framework to systematically assess the performance of GPT-4 on seven categories of radiology tasks (sentence similarity, natural language inference, disease classification, entity extraction, disease progression, finding summary), and compare it with domain-specific SOTA models and GPT-3.5.
Exploring the Impact of Model Scaling on Parameter-Efficient Tuning
Yusheng Su (Tsinghua University), Maosong Sun (Tsinghua University)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Investigate the impact of model scale expansion on performance differences of parameter-efficient tuning (PET) methods, and propose a general PET framework APET that can insert parameters at any position for systematic ablation experiments
Expository Text Generation: Imitate, Retrieve, Paraphrase
Nishant Balepur (University of Illinois at Urbana-Champaign), Kevin Chang
GenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Propose and implement a framework named IRP (Imitate, Retrieve, Paraphrase) for automatically generating accurate and stylistically consistent expository texts, defining the task as generating multi-sentence expository texts given a topic and a fact corpus.
FACTIFY3M: A benchmark for multimodal fact verification with explainability through 5W Question-Answering
Megha Chakraborty (University of South Carolina), Amitava Das (University of South Carolina)
Explainability and InterpretabilityLarge Language ModelDiffusion modelMultimodalityBenchmark
🎯 What it does: Developed a multimodal fact verification dataset named FACTIFY 3M with 3 million entries, and proposed an interpretable verification method based on 5W QA.
FactKB: Generalizable Factuality Evaluation using Language Models Enhanced with Factual Knowledge
Shangbin Feng (University of Washington), Yulia Tsvetkov (University of Washington)
ClassificationRepresentation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose a knowledge-based pre-training method that first performs fact-based pre-training on language models and then conducts fact consistency evaluation.
FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation
Sewon Min (University of Washington), Hannaneh Hajishirzi (University of Washington)
GenerationLarge Language ModelTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Evaluate the factual accuracy of large language models in long-text generation, propose the FACTSCORE metric, and provide an automated estimation method.
Failures Pave the Way: Enhancing Large Language Models through Tuning-free Rule Accumulation
Zeyuan Yang (Tsinghua University), Yang Liu (Tsinghua University)
Data-Centric LearningTransformerLarge Language ModelPrompt EngineeringTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the TRAN framework, enabling models to learn from errors and avoid repeated mistakes in an online learning environment without parameter tuning for large language models (LLMs), through rule generation, evaluation, and accumulation.
Fair Text Classification with Wasserstein Independence
Thibaud Leteno (University of Saint Etienne), Christophe Gravier (University of Saint Etienne)
ClassificationTransformerLarge Language ModelGenerative Adversarial NetworkText
🎯 What it does: Propose an unsupervised fair text classification method (WFC) based on Wasserstein mutual information minimization, which enforces independence between the target and sensitive attributes in the latent space through a pre-trained sensitive attribute model and Wasserstein regularization, without requiring access to sensitive labels during training or inference.
Fair Without Leveling Down: A New Intersectional Fairness Definition
Gaurav Maheshwari (University of Lille), Mikaela Keller (University of Lille)
Convolutional Neural NetworkTransformerImageText
🎯 What it does: Proposed a novel cross-group fairness metric (α-Intersectional Fairness, IFα), and used it to evaluate existing fairness methods, revealing the common 'leveling down' phenomenon.
Faithful Model Evaluation for Model-Based Metrics
Qian Hu (Amazon Alexa AI), Rahul Gupta (Amazon Alexa AI)
TextBenchmark
🎯 What it does: This paper studies methods for conducting statistical significance testing when using model-based metrics, and proposes a framework for estimating variance that includes model prediction errors;
FAME: Flexible, Scalable Analogy Mappings Engine
Shahar Jacob (Hebrew University of Jerusalem), Dafna Shahaf (Hebrew University of Jerusalem)
Explainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Proposes FAME, a system that performs analogy mapping using only entity names, automatically extracts common-sense relationships, and generates explainable mappings.
FaMeSumm: Investigating and Improving Faithfulness of Medical Summarization
Nan Zhang (Pennsylvania State University), Rui Zhang (Pennsylvania State University)
GenerationTransformerSupervised Fine-TuningContrastive LearningTextBiomedical Data
🎯 What it does: Investigated the faithfulness issue in medical text summarization and proposed the FAMESUMM framework, enhancing summary faithfulness through contrastive learning and medical knowledge integration.
FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions
Hyunwoo Kim (Allen Institute for Artificial Intelligence), Maarten Sap (Allen Institute for Artificial Intelligence)
TransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkChain-of-Thought
🎯 What it does: Designed and constructed the FANTOM benchmark to evaluate large language models' theoretical theory of mind (ToM) capabilities in multi-party dialogues with information asymmetry, and conducted rigorous testing through multiple question formats (belief, answerability, info-access).
Fast and Accurate Factual Inconsistency Detection Over Long Documents
Barrett Lattimer (ASAPP), Yi Yang (ASAPP)
Anomaly DetectionExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed the SCALE method based on NLI, which detects factual inconsistencies in generated text by splitting long texts into chunks and reasoning between each chunk and generated sentences.
Fast and Robust Early-Exiting Framework for Autoregressive Language Models with Synchronized Parallel Decoding
Sangmin Bae (KAIST AI), Se-Young Yun (KAIST AI)
Computational EfficiencyKnowledge DistillationTransformerText
🎯 What it does: Propose a Fast and Robust Early-Exiting (FREE) framework to accelerate the inference speed of autoregressive language models;
Faster Minimum Bayes Risk Decoding with Confidence-based Pruning
Julius Cheng (University of Cambridge), Andreas Vlachos (University of Cambridge)
Computational EfficiencyText
🎯 What it does: Proposed an iterative confidence pruning algorithm to accelerate Minimum Bayes Risk (MBR) decoding and reduce the number of utility function calls.
Federated Learning of Large Language Models with Parameter-Efficient Prompt Tuning and Adaptive Optimization
Tianshi Che (Auburn University), Dejing Dou (Baidu Inc)
Federated LearningSafty and PrivacyLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposes the FedPepTAO framework in a federated learning environment for parameter-efficient prompt tuning of large language models, enhancing performance through adaptive optimization.
Federated Meta-Learning for Emotion and Sentiment Aware Multi-modal Complaint Identification
Apoorva Singh (IIT Patna), Tanmay Sen (Ericsson)
ClassificationFederated LearningMeta LearningConvolutional Neural NetworkRecurrent Neural NetworkTransformerMultimodality
🎯 What it does: Under a multimodal (text + image) setting, a federated meta-learning framework is constructed to identify consumer complaints using sentiment and emotion labels as auxiliary information.
FedID: Federated Interactive Distillation for Large-Scale Pretraining Language Models
Xinge Ma (Yunnan University), Xuejie Zhang (Yunnan University)
Federated LearningKnowledge DistillationTransformerLarge Language ModelTextBenchmark
🎯 What it does: Propose the Federated Interactive Distillation (FedID) framework, which uses a small amount of labeled data on the server to correct confirmation bias in traditional federated distillation, achieving decentralized training of large-scale pre-trained language models
FedTherapist: Mental Health Monitoring with User-Generated Linguistic Expressions on Smartphones via Federated Learning
Jaemin Shin (KAIST), Sung-Ju Lee (KAIST)
ClassificationFederated LearningSafty and PrivacyComputational EfficiencyKnowledge DistillationTransformerText
🎯 What it does: Proposes FedTherapist, a mobile mental health monitoring system based on federated learning, utilizing user-generated speech transcriptions and keyboard input text data;