Annual Meeting of the Association for Computational Linguistics · 518 papers
Hierarchical Attention Generates Better Proofs
Jianlong Chen (Chinese University of Hong Kong), Andrew C Yao (Shanghai Qi Zhi Institute)
CodeAI Code AssistantTransformerLarge Language ModelTextBenchmark
🎯 What it does: Proposed a new regularization method called Hierarchical Attention, aiming to improve the performance of large language models (LLMs) in mathematical proofs by establishing a five-level hierarchy to guide the model's attention mechanism.
Hierarchical Memory Organization for Wikipedia Generation
Eugene J. Yu (Peking University), Sujian Li (Peking University)
CodeGenerationTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Proposed a Wikipedia automatic generation framework MOG based on hierarchical memory structure, which can refine massive web information into factual units and organize them according to Wiki structure, ultimately generating complete and traceable entries.
HoH: A Dynamic Benchmark for Evaluating the Impact of Outdated Information on Retrieval-Augmented Generation
Jie Ouyang (University of Science and Technology of China), Qi Liu (University of Science and Technology of China)
CodeRetrievalLarge Language ModelTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Designed and implemented the HOH benchmark for systematically evaluating the performance of Retrieval-Augmented Generation (RAG) when facing outdated information, and automatically generated large-scale dynamic QA datasets and simulated search engines through token-level diff and LLM pipelines;
HomeBench: Evaluating LLMs in Smart Homes with Valid and Invalid Instructions Across Single and Multiple Devices
Silin Li (Beijing Institute of Technology), Haifeng Wang (Baidu Inc)
CodeTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed the HomeBench dataset to evaluate the ability of LLMs to handle valid/invalid single-device and multi-device instructions in smart home environments, constructed a customizable virtual home environment and generated approximately 170,000 instructions;
🎯 What it does: Constructed MISBENCH, a large-scale and diverse misinformation benchmark, to evaluate the performance of large language models (LLMs) under different types of knowledge conflicts and text styles; simultaneously proposed the RtD method based on retrieval and reconstruction to enhance LLMs' ability to identify misinformation.
How to Compare Things Properly? A Study of Argument Relevance in Comparative Question Answering
Irina Nikishina (University of Hamburg), Chris Biemann (University of Hamburg)
CodeGenerationRetrievalTransformerLarge Language ModelPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: Constructed a manually annotated comparative argument relevance dataset, and generated answers for comparison questions based on this dataset (the silver dataset was automatically generated by ChatGPT, while the gold dataset was manually refined by human editors).
HSCR: Hierarchical Self-Contrastive Rewarding for Aligning Medical Vision Language Models
Songtao Jiang (Zhejiang University), Zuozhu Liu (Zhejiang University)
CodeReinforcement Learning from Human FeedbackTransformerSupervised Fine-TuningVision Language ModelContrastive LearningBiomedical Data
🎯 What it does: This paper proposes a Hierarchical Self-Contrastive Rewarding (HSCR) method, which generates high-quality preference data through self-contrastive rewarding and optimizes preferences at multiple levels, thereby enhancing modality alignment and reliability in medical vision-language models (Med-VLM).
HumT DumT: Measuring and controlling human-like language in LLMs
Myra Cheng (Stanford University), Dan Jurafsky (Stanford University)
CodeGenerationOptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: This paper proposes two metrics based on LLM probabilities, HUMT and SOCIOT, to quantify the human-like tone in generated text, and employs the DUMT method to reduce the level of human-likeness while maintaining model performance.
HyKGE: A Hypothesis Knowledge Graph Enhanced RAG Framework for Accurate and Reliable Medical LLMs Responses
Xinke Jiang (Peking University), Yasha Wang (Peking University)
CodeTransformerLarge Language ModelPrompt EngineeringTextGraphRetrieval-Augmented Generation
🎯 What it does: A retrieval-augmented generation framework called HyKGE based on a Hypothesis Knowledge Graph (HKG) was constructed to enhance the accuracy and reliability of large language models in the medical domain.
I0T: Embedding Standardization Method Towards Zero Modality Gap
Na Min An (KAIST), Hyunjung Shim (KAIST)
CodeRetrievalRepresentation LearningTransformerVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: Propose and study a framework named I0T, aiming to significantly reduce the modality gap between image and text embeddings in CLIP models through methods such as embedding normalization or batch normalization.
IAM: Efficient Inference through Attention Mapping between Different-scale LLMs
Yi Zhao (Shanghai Jiao Tong University), Hai Zhao (Shanghai Jiao Tong University)
CodeComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Leverage the attention matrix of a small LLM to map to a large LLM, accelerating inference and reducing KV cache usage while maintaining model performance.
Identifying Cellular Niches in Spatial Transcriptomics: An Investigation into the Capabilities of Large Language Models
Huanhuan Wei (Yale School of Medicine), Xiting Yan (Yale School of Medicine)
CodeRecognitionExplainability and InterpretabilityData-Centric LearningGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringBiomedical Data
🎯 What it does: Leverage large language models (LLMs) for spatial environment identification in spatial transcriptomics data, proposing the LLMiniST method with zero-shot and two-stage fine-tuning to automatically generate interpretable spatial niche labels.
Identifying Reliable Evaluation Metrics for Scientific Text Revision
Leane Jourdan, Richard Dufour (Nantes Université)
CodeTransformerLarge Language ModelTextBenchmark
🎯 What it does: Study how to reliably evaluate the quality of scientific text revisions, systematically assess the effectiveness of traditional similarity metrics, cross-domain metrics, and LLM-as-a-judge methods.
🎯 What it does: Constructed a video dataset named ImpliHateVid for implicit hate speech detection, and proposed a two-stage contrastive learning framework to identify implicit hate content in videos.
Improve Safety Training of Large Language Models with Safety-Critical Singular Vectors Localization
Peijian Gu (University Of Science And Technology Of China), Zhendong Mao (University Of Science And Technology Of China)
CodeSafty and PrivacyComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: The study proposes a pluggable method that initializes LoRA by locating safety-critical singular vectors in model parameters, enabling safe training to update only safety-related parameters and minimizing impact on the model's general functionality.
CodeDrug DiscoveryTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringBiomedical Data
🎯 What it does: This paper explores the feasibility of using large language models (LLMs) as judges (LLM-Judge) to automatically evaluate biomedical relation extraction models, and proposes structured output formats and domain adaptation techniques to improve judgment accuracy.
🎯 What it does: Proposed the RHIO framework, which generates realistic untrustworthy samples through masked retrieval heads, trains models to distinguish between trustworthy and untrustworthy answers using control codes, and further enhances contextual credibility during inference via self-contrast decoding.
🎯 What it does: Proposed CombiSearch, a retrieval example method based on combinatorial search for dialogue state tracking tasks, which selects examples that synergistically improve model performance during retrieval.
CodeGenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Designed and implemented the Ewe system, introducing explicit working memory in long-text generation, which receives real-time retrieval and fact-checking feedback, periodically pauses generation, refreshes memory, and corrects errors;
Improving Fairness of Large Language Models in Multi-document Summarization
Haoyuan Li (University of North Carolina at Chapel Hill), Snigdha Chaturvedi (University of North Carolina at Chapel Hill)
CodeOptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: This paper proposes FairPO, a preference optimization method for multi-document summarization, aiming to simultaneously improve fairness at both the summary level and the corpus level.
🎯 What it does: Developed the BELOPSEM benchmark and improved and evaluated unsupervised sentence alignment and clustering-based isometric enhancement for sentence mining on three pairs of low-resource/endangered language pairs.
🎯 What it does: This paper created the IndicSynth dataset, containing approximately 4000 hours of multilingual synthetic speech (12 low-resource Indian languages) from 989 speakers, divided into two subsets: realistic mimicry and diversity.
Inferring Functionality of Attention Heads from their Parameters
Amit Elhelo (Tel Aviv University), Mor Geva (Tel Aviv University)
CodeExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose the MAPS framework, which directly infers the functionality of attention heads using their parameters without requiring training or inference.
Innovative Image Fraud Detection with Cross-Sample Anomaly Analysis: The Power of LLMs
QiWen Wang, Chen Lin (Xiamen University)
CodeAnomaly DetectionTransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmarkFinance RelatedRetrieval-Augmented Generation
🎯 What it does: Propose the CSIAD framework for image fraud detection based on cross-sample logical reasoning, combining retrieval, rule generation, and fact verification to accurately locate fine-grained forged regions and provide explanations.
Instance-Selection-Inspired Undersampling Strategies for Bias Reduction in Small and Large Language Models for Binary Text Classification
Guilherme Fonseca (Universidade Federal de Minas Gerais), Leonardo Chaves Dutra Da Rocha (Universidade Federal de São João del Rei)
CodeClassificationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper addresses the bias caused by class imbalance in binary text classification tasks, focusing on small and large language models (RoBERTa, Llama3.1), and proposes two instance-based under-sampling methods (E2SC_US and UBR). These methods are compared with 19 baseline approaches across 13 datasets.
INTERACT: Enabling Interactive, Question-Driven Learning in Large Language Models
Aum Kendapadi (University Of North Carolina Chapel Hill), Shashank Srivastava (University Of North Carolina Chapel Hill)
CodeKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Proposes the INTERACT framework, enabling large language models (LLM) to undergo interactive, question-driven learning through student-teacher dialogues;
Interlocking-free Selective Rationalization Through Genetic-based Learning
Federico Ruggeri (University of Bologna), Gaetano Signorelli (University of Bologna)
CodeClassificationData SynthesisExplainability and InterpretabilityRecurrent Neural NetworkText
🎯 What it does: Propose the GenSPP framework, which uses genetic algorithms to separate the training of the generator and predictor, achieving unsupervised selective rationalization and completely eliminating the interlocking problem in traditional select-then-predict models.
Internal and External Impacts of Natural Language Processing Papers
Yu Zhang (Texas A&M University)
CodeText
🎯 What it does: This paper conducts a large-scale scientometric analysis of academic internal citations (OpenAlex) and external citations (patents, media, policy documents) of papers published in ACL, EMNLP, and NAACL from 1979 to 2024, quantifying their influence across different fields based on paper topics.
Internal Value Alignment in Large Language Models through Controlled Value Vector Activation
Haoran Jin (University of Science and Technology of China), Defu Lian (University of Science and Technology of China)
CodeExplainability and InterpretabilityAdversarial AttackTransformerLarge Language ModelContrastive LearningText
🎯 What it does: This paper proposes the ConVA framework, which achieves fine-grained alignment of model values by identifying and activating value vectors within LLMs;
Investigating and Extending Homans’ Social Exchange Theory with Large Language Model based Agents
Lei Wang (Renmin University of China), Xu Chen (Renmin University of China)
CodeLarge Language ModelAgentic AIText
🎯 What it does: This paper utilizes large language model (LLM) agents to construct a virtual society, conducts social exchange games, and verifies and extends Homans' social exchange theory.
iQUEST: An Iterative Question-Guided Framework for Knowledge Base Question Answering
Shuai Wang (Chalmers University of Technology), Yinan Yu (Chalmers University of Technology)
CodeGraph Neural NetworkTransformerLarge Language ModelGraphBenchmarkChain-of-Thought
🎯 What it does: Proposed the iQUEST framework, which achieves multi-hop knowledge graph reasoning by iteratively generating sub-questions and using graph neural networks for two-hop neighbor retrieval.
Is linguistically-motivated data augmentation worth it?
Ray Groshan (University of Colorado), Alexis Palmer (University of Colorado)
CodeRecognitionGenerationTransformerText
🎯 What it does: This paper systematically compares two categories of data augmentation methods for low-resource languages, evaluating their effectiveness on machine translation and linear annotation tasks.
Is That Your Final Answer? Test-Time Scaling Improves Selective Question Answering
William Jurayj, Benjamin Van Durme (Johns Hopkins University)
CodeComputational EfficiencyLarge Language ModelTextChain-of-Thought
🎯 What it does: Explore introducing confidence thresholds when scaling large language models at test time, enabling selective answering and evaluating their impact on confidence and computational budget.
ISR: Self-Refining Referring Expressions for Entity Grounding
Zhuocheng Yu (Peking University), Zhonghui He (Peking University)
CodeRecognitionRetrievalTransformerLarge Language ModelReinforcement LearningVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes an Iterative Self-Optimization (ISR) training framework that leverages multimodal large language models (MLLM) to automatically generate and iteratively refine referential expressions (RE), thereby enhancing the performance of the entity grounding task.
It’s Not Bragging If You Can Back It Up: Can LLMs Understand Braggings?
Jingjie Zeng (Dalian University of Technology), Hongfei Lin (Dalian University of Technology)
CodeRecognitionGenerationTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: This paper systematically studies the understanding and generation of self-praise behavior in large language models (LLMs), constructing three tasks (identification, explanation, generation) and designing corresponding evaluation metrics.
Jailbreak Large Vision-Language Models Through Multi-Modal Linkage
Yu Wang (Chinese Academy of Sciences), Tianxing He (Tsinghua University)
CodeAdversarial AttackPrompt EngineeringVision Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Propose a Multimodal Link (MML) framework that achieves exploitation on vision-language models by encrypting malicious text in images and guiding VLMs to decrypt it.
JailbreakRadar: Comprehensive Assessment of Jailbreak Attacks Against LLMs
Junjie Chu (CISPA Helmholtz Center for Information Security), Yang Zhang (CISPA Helmholtz Center for Information Security)
CodeSafty and PrivacyAdversarial AttackTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This paper systematically evaluates 17 representative jailbreak attacks, proposes a six-class attack taxonomy, constructs a unified prohibited questions dataset with 160 questions and 16 violation categories, and assesses attack effectiveness and multiple defense methods on nine alignment LLMs, revealing attack patterns and defense challenges.
Just Go Parallel: Improving the Multilingual Capabilities of Large Language Models
Muhammad Reza Qorib (National University of Singapore), Hwee Tou Ng (National University of Singapore)
CodeTransformerLarge Language ModelText
🎯 What it does: The study systematically evaluates and quantifies the role of parallel data in enhancing the multilingual capabilities of large language models, focusing on translation and multilingual commonsense reasoning.
K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Korean
Minkyeong Jeon (Korea University), Byung-Jun Lee (Korea University)
CodeGenerationData SynthesisLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: Proposes K/DA—a pipeline for automatically generating paired Korean offensive language datasets, focusing on trending slang and implicit aggression;
KiRAG: Knowledge-Driven Iterative Retriever for Enhancing Retrieval-Augmented Generation
Jinyuan Fang (University of Glasgow), Craig MacDonald
CodeGenerationRetrievalLarge Language ModelContrastive LearningTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the KiRAG model, which enhances multi-hop question answering by splitting documents into knowledge triplets, performing knowledge-driven iterative retrieval, sorting documents using retrieved triplets, and generating answers with LLMs, thereby improving both retrieval and generation effectiveness.
Know You First and Be You Better: Modeling Human-Like User Simulators via Implicit Profiles
Kuang Wang (Chinese University of Hong Kong), Haizhou Li (Chinese University of Hong Kong)
CodeReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningContrastive LearningText
🎯 What it does: This paper proposes a user simulation framework called USP based on implicit user profiling, which can extract latent features from human-computer dialogues and generate personalized, coherent user responses.
Knowledge Boundary of Large Language Models: A Survey
Moxin Li (National University Of Singapore), Yang Deng (Singapore Management University)
CodeTransformerLarge Language ModelPrompt EngineeringTextReview/Survey PaperRetrieval-Augmented Generation
🎯 What it does: A systematic review of the knowledge boundaries of large language models (LLMs), proposing a unified formal definition and four types of knowledge classification (universal knowledge boundaries, parameterized knowledge boundaries, outward knowledge boundaries), and summarizing and commenting on existing methods and datasets around three research questions: 'why study', 'how to identify', and 'how to alleviate'.
Knowledge Tracing in Programming Education Integrating Students’ Questions
Doyoun Kim (Seoul National University), Yohan Jo (Seoul National University)
CodeTransformerLarge Language ModelContrastive LearningTextMultimodality
🎯 What it does: This study proposes SQKT, a knowledge tracking model that integrates student questions and automatically extracts skill information to predict students' learning outcomes on subsequent problems in programming courses.
Knowledge-Augmented Multimodal Clinical Rationale Generation for Disease Diagnosis with Small Language Models
Shuai Niu (Hong Kong Baptist University), Xian Yang (University of Manchester)
CodeExplainability and InterpretabilityKnowledge DistillationTransformerLarge Language ModelTextMultimodalityTime SeriesElectronic Health RecordsRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Generate multimodal (text + time-series lab tests) disease diagnosis and interpretable reasoning (clinical rationale) using small language models (SLM) through knowledge-enhanced attention mechanisms and sequential reasoning distillation.
KnowShiftQA: How Robust are RAG Systems when Textbook Knowledge Shifts in K-12 Education?
Tianshi Zheng (Hong Kong University of Science and Technology), Yangqiu Song (University of Tokyo)
CodeRetrievalTransformerLarge Language ModelPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: This paper constructs a K-12 education QA dataset named KNOWSHIFTQA to evaluate the robustness of Retrieval-Augmented Generation (RAG) systems in scenarios involving changes to textbook knowledge.
KV-Latent: Dimensional-level KV Cache Reduction with Frequency-aware Rotary Positional Embedding
Shi Luohe (Wuhan University), Hai Zhao (Wuhan University)
CodeComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose the KV-Latent framework, which significantly compresses the KV cache and accelerates inference by downsampling the Key/Value dimensions of Transformer attention heads into a latent space; meanwhile, performance is restored through two-stage training (layer-wise distillation + end-to-end fine-tuning).
L4Q: Parameter Efficient Quantization-Aware Fine-Tuning on Large Language Models
Hyesung Jeon (Seoul National University), Jae-Joon Kim (Seoul National University)
CodeComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed L4Q, a parameter-efficient fine-tuning method that combines quantization-aware training (QAT) with LoRA, achieving fully quantized and memory-efficient LLM fine-tuning.
LAMB: A Training-Free Method to Enhance the Long-Context Understanding of SSMs via Attention-Guided Token Filtering
Zhifan Ye (Georgia Institute of Technology), Souvik Kundu (Intel Labs)
CodeContrastive LearningTextBenchmark
🎯 What it does: Propose a training-free attention-guided token filtering method called LAMB to enhance the long-context understanding capability of state-space model (SSM)-based models.
Language Models can Subtly Deceive Without Lying: A Case Study on Strategic Phrasing in Legislation
Atharvan Dogra (Indian Institute of Technology Madras), Balaraman Ravindran (Indian Institute of Technology Madras)
CodeTransformerLarge Language ModelPrompt EngineeringTextFinance Related
🎯 What it does: Established a testing platform for legislative environments to investigate how LLMs can conceal deceptive actions favoring specific companies through subtle wording.
LAQuer: Localized Attribution Queries in Content-grounded Generation
Eran Hirsch (Bar-Ilan University), Ido Dagan (Bar-Ilan University)
CodeGenerationExplainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Proposed the LAQuer task, allowing users to query and locate corresponding source document fragments after selecting segments in the generated text, achieving localized attribution;
CodeDrug DiscoveryGraph Neural NetworkTransformerLarge Language ModelContrastive LearningTextGraphBiomedical DataRetrieval-Augmented Generation
🎯 What it does: Propose the LLaPA model for protein-protein interaction (PPI) prediction, integrating protein sequences and PPI network information into a large language model to predict multi-label PPI types and multi-protein complex affinity.
🎯 What it does: Proposes the MARAL framework, combining maximum margin contrastive learning with progressive cross-lingual adaptation techniques to enhance named entity recognition performance for low-resource languages.
Yanyang Li (Chinese University of Hong Kong), Liwei Wang (Chinese University of Hong Kong)
CodeOptimizationReinforcement Learning from Human FeedbackText
🎯 What it does: Proposed a paradigm called FTTT that utilizes feedback for optimization during testing, and designed a learnable test-time optimizer named OPTUNE;
Learning to Reason Over Time: Timeline Self-Reflection for Improved Temporal Reasoning in Language Models
Adrián Bazaga (University of Cambridge), Adrià de Gispert (Amazon AGI)
CodeExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextChain-of-Thought
🎯 What it does: Proposes the Temporal Self-Reflective Prompting (TISER) framework, dividing the LLM reasoning process into four stages (initial reasoning → timeline construction → self-reflection → answer generation), and extending the reasoning chain through test-time scaling; meanwhile, constructs a synthetic dataset containing intermediate reasoning trajectories for model fine-tuning.
Learning to Rewrite: Generalized LLM-Generated Text Detection
Wei Hao (Columbia University), Chengzhi Mao (Rutgers University)
CodeAnomaly DetectionTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Propose the L2R (Learning to Rewrite) framework, which fine-tunes LLMs to generate extensive edits when rewriting human text, while making minimal changes to LLM-generated text, creating a clear distinction in edit distance for detecting LLM-generated text.
Yifei Yang (Shanghai Jiao Tong University), Hai Zhao (Central South University)
CodeComputational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposed a learnable LLM layer depth expansion method called LESA, which achieves model depth expansion by predicting intermediate layer parameters between adjacent layers.
🎯 What it does: The paper proposes a vocabulary diversity-aware retrieval-augmented generation framework called DRAG, which enhances the factual accuracy of large language models (LLMs) by leveraging fine-grained query splitting, relevance assessment, and risk-guided sparse correction.
Lexical Recall or Logical Reasoning: Probing the Limits of Reasoning Abilities in Large Language Models
Henrike Beyer (University of Dundee), Chris Reed (University of Dundee)
CodeTransformerLarge Language ModelTextBenchmark
🎯 What it does: Constructed and evaluated a logic grid puzzle benchmark named Mystery-Zebra, containing 4,290 puzzles, designed to test LLMs' reasoning capabilities under lexical interference and formal difficulty levels.
🎯 What it does: Propose a three-stage legal text generation framework called LexKeyPlan, which first generates legal concept key phrases as a prospective content plan, then uses this plan to retrieve external documents, and finally generates legal reasoning text.
Library-Like Behavior In Language Models is Enhanced by Self-Referencing Causal Cycles
Munachiso S Nwadike, Kentaro Inui (Mohamed bin Zayed University of Artificial Intelligence)
CodeTransformerLarge Language ModelPrompt EngineeringTextSequentialRetrieval-Augmented Generation
🎯 What it does: Proposed and verified the self-referential causal loop (RECALL) mechanism, utilizing cycle tokens to enable large language models (LLMs) to transcend sequential limitations and overcome the 'reversal curse'.
Linguistic Generalizability of Test-Time Scaling in Mathematical Reasoning
Guijin Son (Yonsei University), James Thorne (KAIST AI)
CodeComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark
🎯 What it does: This paper investigates the language generalization of three test-time scaling methods (Outcome Reward Modeling, Process Reward Modeling, Budget Forcing) on competition-level math reasoning tasks across 55 languages, and constructs and releases the multilingual competition-level math benchmark MCLM as well as the multilingual reasoning LLM MR1-1.5B.
Literary Evidence Retrieval via Long-Context Language Models
Katherine Thai (UMass Amherst), Mohit Iyyer (University of Maryland)
CodeGenerationRetrievalTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Explored how to leverage long-context large language models (LLMs) to complete literary evidence retrieval tasks, i.e., automatically generating missing citations given complete novel texts and missing literary criticism fragments.
Qingkai Fang (Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences), Yang Feng (Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences)
CodeGenerationData SynthesisTransformerLarge Language ModelFlow-based ModelTextAudio
🎯 What it does: Propose the LLaMA-Omni 2 speech-language model to enable real-time high-quality voice interaction.
LlamaDuo: LLMOps Pipeline for Seamless Migration from Service LLMs to Small-Scale Local LLMs
Chansung Park (Electronics and Telecommunications Research Institute), Jing Tang (Hong Kong University of Science and Technology)
CodeData SynthesisComputational EfficiencyKnowledge DistillationData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed LlamaDuo, an automated LLMOps workflow for transferring knowledge from cloud-based service LLMs to smaller, locally deployable LLMs, ensuring service continuity in offline or restricted environments.
LLäMmlein: Transparent, Compact and Competitive German-Only Language Models from Scratch
Jan Pfister (Julius-Maximilians-Universität Würzburg), Andreas Hotho (Julius-Maximilians-Universität Würzburg)
CodeGenerationRepresentation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Train and publicly release two full German decoder language models, LLäMmlein 120M and 1B, from scratch, providing complete training data, code, distributed training details, iterative checkpoints, and system evaluations, aiming to establish a reproducible and transparent baseline for the German NLP research community.
LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering
Jinhe Bi (Ludwig Maximilian University of Munich), Yunpu Ma (Ludwig Maximilian University of Munich)
CodeComputational EfficiencyRepresentation LearningTransformerSupervised Fine-TuningVision Language ModelMultimodalityBenchmark
🎯 What it does: This paper proposes the Modality Linear Representation-Steering (MoReS) method, which rebalances visual and textual modalities during the visual instruction tuning phase by applying linear subspace mapping to visual representations, and builds the LLaVA Steering model along with a visualization evaluation platform based on this.
Georg Wölflein (EKFZ for Digital Health TU Dresden), Jakob Nikolas Kather (EKFZ for Digital Health TU Dresden)
CodeAI Code AssistantTransformerLarge Language ModelAgentic AITextBenchmark
🎯 What it does: This study proposes the TOOLMAKER framework, which can automatically convert GitHub code repositories corresponding to scientific papers into tools callable by large language models (LLMs), and evaluates its performance through the TM-BENCH benchmark.
LLM as Entity Disambiguator for Biomedical Entity-Linking
Christophe Ye (Georgia Institute of Technology), Cassie S. Mitchell (Georgia Institute of Technology)
CodeRetrievalTransformerLarge Language ModelBiomedical DataRetrieval-Augmented Generation
🎯 What it does: In the biomedical entity linking task, this paper proposes using a large language model (LLM) as the core entity disambiguator, directly embedding it into the subsequent disambiguation steps after candidate generation, thereby improving linking accuracy without any fine-tuning.
LLM Meets Scene Graph: Can Large Language Models Understand and Generate Scene Graphs? A Benchmark and Empirical Study
Dongil Yang (Yonsei University), Jinyoung Yeo (Yonsei University)
CodeGenerationData SynthesisTransformerLarge Language ModelTextGraphBenchmarkChain-of-Thought
🎯 What it does: Proposes a scene graph benchmark (TSG Bench) tailored for large language models (LLMs) to evaluate their capabilities in two major tasks: scene graph understanding and generation.
LLM-Guided Semantic-Aware Clustering for Topic Modeling
Jianghan Liu (Southeast University), Yining Li (Southeast University)
CodeTransformerLarge Language ModelText
🎯 What it does: This paper proposes LiSA, a topic model that combines large language models (LLMs) with clustering methods. It first generates candidate topic words for each document using LLMs, then performs clustering on both documents and candidate topic words. It utilizes LLMs for mapping and validation, and aligns the document-level and topic-level semantic spaces globally through a collaborative enhancement phase, ultimately achieving more accurate topic distributions.
LLM-Powered Test Case Generation for Detecting Bugs in Plausible Programs
Kaibo Liu (Peking University), Gang Huang (Peking University)
CodeGenerationAI Code AssistantTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes TrickCatcher, a test case generation method based on large language models, for detecting tricky bugs in trustworthy programs.
Jiongnan Liu (Renmin University of China), Zhicheng Dou (Renmin University of China)
CodeRecommendation SystemComputational EfficiencyRepresentation LearningData-Centric LearningTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Proposed a lightweight, pluggable personalized LLM method called PPlug, achieving parameter-free personalized generation by encoding user historical behavior into a single user embedding and appending it to the LLM input.
LLMs Can Simulate Standardized Patients via Agent Coevolution
Zhuoyun Du (Zhejiang University), Haochao Ying (Sun Yat-sen University)
CodeTransformerLarge Language ModelAgentic AITextBiomedical DataElectronic Health RecordsRetrieval-Augmented Generation
🎯 What it does: This paper proposes EvoPatient, a multi-agent co-evolution framework that automatically simulates and records dialogues between patient and doctor agents generated by LLMs during simulated medical consultations. Through self-supervised mechanisms, it continuously enhances the standardization of patient presentations and the professionalism of doctors' questioning, achieving standardized patient (SP) simulation and physician training without human supervision.
LLMs Caught in the Crossfire: Malware Requests and Jailbreak Challenges
Haoyang Li (Institute of Artificial Intelligence (TeleAI) China Telecom), Xuelong Li (Institute of Artificial Intelligence (TeleAI) China Telecom)
CodeSafty and PrivacyTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This study constructs the MalwareBench benchmark dataset and evaluates the security performance of 29 mainstream large language models (LLMs) under malicious code generation and black-box jailbreak attacks.
LoGU: Long-form Generation with Uncertainty Expressions
Ruihan Yang (Fudan University), Deqing Yang (Fudan University)
CodeGenerationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Investigated how large language models express uncertainty in long-text generation and proposed the LoGU (Long-form Generation with Uncertainty) task;
LongBench v2: Towards Deeper Understanding and Reasoning on Realistic Long-context Multitasks
Yushi Bai (Tsinghua University), Juanzi Li (Tsinghua University)
CodeTransformerLarge Language ModelTextBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose LongBench v2, a multi-task benchmark for long-text understanding and reasoning, containing 503 questions with lengths ranging from 8k to 2M words, covering six major task categories.
Lost in Literalism: How Supervised Training Shapes Translationese in LLMs
Yafu Li (Shanghai AI Laboratory), Yue Zhang (Westlake University)
CodeGenerationData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Systematically evaluate over-literal translation (translationese) in LLM translations and investigate its root causes from the supervised fine-tuning (SFT) stage.
LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion
Guanghao Zhou (East China Normal University), Jun Zhou (Ant Group)
CodeSafty and PrivacyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose a post-hoc safety realignment framework LSSF, which restores compromised safety after fine-tuning through low-rank safety subspace fusion.
Jiaheng Liu (Nanjing University), Bo Zheng (Alibaba Group)
CodeAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Constructed a repository-level code completion benchmark, MRC-EVAL 2, covering 18 mainstream programming languages, providing two fine-grained annotations (bucket-level and semantic-level), while organizing a multilingual instruction corpus, MRC-INSTRUCT 2, to enhance the completion capabilities of existing code LLMs.
M³GQA: A Multi-Entity Multi-Hop Multi-Setting Graph Question Answering Benchmark
Boci Peng (Peking University), Yan Zhang (Peking University)
CodeTransformerLarge Language ModelGraphBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed and constructed M3GQA, a multi-entity, multi-hop, multi-setting graph question answering benchmark containing diverse queries, answers, and semantically correct reasoning paths.
Make Imagination Clearer! Stable Diffusion-based Visual Imagination for Multimodal Machine Translation
Andong Chen (Harbin Institute of Technology), Min Zhang (Harbin Institute of Technology)
CodeImage TranslationGenerationTransformerLarge Language ModelReinforcement LearningVision Language ModelDiffusion modelImageTextMultimodality
🎯 What it does: Propose the IMAGE framework, which generates images from source sentences using Stable Diffusion and then translates them using a multimodal LLM.
Making FETCH! Happen: Finding Emergent Dog Whistles Through Common Habitats
Kuleen Sasse (Johns Hopkins University), Mark Dredze (Johns Hopkins University)
CodeClassificationTransformerLarge Language ModelPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose the FETCH! task, automatically discovering emerging dog whistles in large-scale social media corpora and constructing a benchmark evaluation system.
Making LLMs Better Many-to-Many Speech-to-Text Translators with Curriculum Learning
Yexing Du (Harbin Institute of Technology), Bing Qin (Harbin Institute of Technology)
CodeTransformerLarge Language ModelSupervised Fine-TuningMultimodalityAudio
🎯 What it does: Reframe the speech-to-text translation (S2TT) task as a speech recognition + translation (SRT) task, leveraging the machine translation capability of large language models through a three-stage curriculum learning to achieve multilingual end-to-end S2TT.
MAPLE: Enhancing Review Generation with Multi-Aspect Prompt LEarning in Explainable Recommendation
Ching-Wen Yang (National Cheng Kung University), Hung-Yu Kao (National Tsing Hua University)
CodeRecommendation SystemExplainability and InterpretabilityTransformerPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: Propose the MAPLE model, which generates personalized, accurate, and diverse recommendation explanations through multi-aspect prompting learning.
Mapping the Podcast Ecosystem with the Structured Podcast Research Corpus
Benjamin Roger Litterer, Dallas Card (University of Michigan)
CodeTextMultimodalityAudio
🎯 What it does: This paper first constructs and publicly releases the SPORC dataset, which includes 1.1 million podcast transcriptions, covering metadata, audio features, speaker separation, and host/guest role annotations. Based on this dataset, the paper conducts a systematic analysis of podcast content, network structures, and responses to major events.
MAPS: Motivation-Aware Personalized Search via LLM-Driven Consultation Alignment
Weicong Qin (Renmin University of China), Jun Xu (Renmin University of China)
CodeRecommendation SystemTransformerLarge Language ModelMixture of ExpertsContrastive LearningText
🎯 What it does: By introducing AI consultation texts, extracting users' search intent, and utilizing LLM with Mixture of Attention Experts (MoAE) to map queries and consultations into a unified semantic space, achieving personalized product retrieval and ranking.
MARS: Benchmarking the Metaphysical Reasoning Abilities of Language Models with a Multi-task Evaluation Dataset
Weiqi Wang (HKUST), Yangqiu Song (HKUST)
CodeTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposed a three-step discriminative process for metaphysical reasoning to evaluate the reasoning capabilities of language models in the face of changes in situational distributions;
🎯 What it does: Propose the MathFusion framework, which combines existing mathematical problems to generate new ones through three fusion strategies: sequential, parallel, and conditional, and further fine-tunes LLMs with instruction-based tuning;
MaXIFE: Multilingual and Cross-lingual Instruction Following Evaluation
Yile Liu (Waseda University), Liang Li (OPPO AI Center)
CodeLarge Language ModelTextBenchmark
🎯 What it does: Proposed and implemented the MaXIFE benchmark for systematic evaluation of multilingual (23 languages) and cross-lingual instruction-following capabilities;
Measuring the Effect of Transcription Noise on Downstream Language Understanding Tasks
Ori Shapira (OriginAI), Amir David Nissan Cohen
CodeClassificationTransformerLarge Language ModelTextAudio
🎯 What it does: Propose a configurable framework (ENDOW) for systematically evaluating the impact of transcription noise on downstream SLU tasks, and conduct experimental analysis on various noise levels, noise types, and cleaning techniques.
MEDDxAgent: A Unified Modular Agent Framework for Explainable Automatic Differential Diagnosis
Daniel Philip Rose (University of California), Carolin Lawrence (NEC Laboratories Europe)
CodeExplainability and InterpretabilityTransformerLarge Language ModelTextBiomedical DataBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed a modular and interpretable automated differential diagnosis framework called MEDDxAgent, which achieves iterative diagnosis through three modules: interactive history acquisition, knowledge retrieval, and diagnostic strategies;
🎯 What it does: Investigated how sequence-level knowledge distillation (SeqKD) enables student models to inherit instance-level memory from teacher models, evaluated changes in memory and hallucination through experiments, and subsequently proposed Adaptive-SeqKD to reduce memory and hallucination via teacher fine-tuning.
Memorizing is Not Enough: Deep Knowledge Injection Through Reasoning
Ruoxi Xu (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences), Le Sun (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences)
CodeTransformerLarge Language ModelTextBenchmarkChain-of-Thought
🎯 What it does: Proposed a four-layer knowledge injection framework (Recall, Extraction, Reasoning, Association) and constructed the DeepKnowledge benchmark to fine-grainedly evaluate the effectiveness of knowledge injection in LLMs;
Meta-Reflection: A Feedback-Free Reflection Learning Framework
Yaoke Wang (Zhejiang University), Yueting Zhuang (Zhejiang University)
CodeMeta LearningAI Code AssistantTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Proposed the Meta-Reflection framework, achieving an LLM improvement scheme that enables reflection without external feedback and through a single inference step.