Papers by Xiangyu Zhang

The Elephant in the Room: Exploring the Role of Neutral Words in Language Model Group-Agnostic Debiasing (2025.findings-acl)

Challenge:	Personalized Large Language Models (PLLMs) aim to align outputs with individual user preferences . current methods of fine-tuning a separate module for each user are unscalable .
Approach:	They propose a Merge-then-Adapt framework for Personalized Large Language Models . they construct a shared Meta-LoRA bank and propose an Adaptive LoRA Fusion stage .
Outcome:	The proposed framework outperforms existing SOTA methods on the LaMP benchmark.

Copied to clipboard

Xinwei Guo, Jiashi Gao, Junlei Zhou, Jiaxin Zhang, Guanhua Chen, Xiangyu Zhao, Quanying Liu, Haiyan Wu, Xin Yao, Xuetao Wei

Hierarchical Memory Organization for Wikipedia Generation (2025.acl-long)

Challenge:	Large Language Models (LLMs) are increasingly integrated into our daily lives, raising ethical concerns, especially about perpetuating stereotypes.
Approach:	They propose a method that incorporates a neutral word semantics-based loss function to alleviate the deterioration of the LMS during debiasing.
Outcome:	The proposed method alleviates the deterioration of the Language Modeling Score (LMS) by incorporating a neutral word semantics-based loss function.

Copied to clipboard

Eugene J. Yu, Dawei Zhu, Yifan Song, Xiangyu Wong, Jiebin Zhang, Wenxuan Shi, Xiaoguang Li, Qun Liu, Sujian Li

Challenge:	Existing methods for generating Wikipedia articles do not utilize memory directly for outline generation.
Approach:	They propose a method to generate Wikipedia articles autonomously by leveraging a hierarchical memory architecture.
Outcome:	The proposed framework outperforms baseline methods in producing informative and reliable articles.

Capturing Event Argument Interaction via A Bi-Directional Entity-Level Recurrent Decoder (2021.acl-long)

Copied to clipboard

Xi Xiangyu, Wei Ye, Shikun Zhang, Quanxiu Wang, Huixing Jiang, Wei Wu

MemoPhishAgent: Memory-Augmented Multi-Modal LLM Agent for Phishing URL Detection (2026.acl-industry)

Challenge:	Existing efforts to capture event argument interactions are limited by the argument role type information of contextual entities.
Approach:	They propose to capture event argument interactions as a Seq2Seq-like learning problem where a sentence with a specific event trigger is mapped to a sequence of event argument roles.
Outcome:	The proposed neural architecture generates argument roles by incorporating contextual entities’ argument role predictions, like a word-by-word text generation process, thereby distinguishing implicit argument distribution patterns within an event more accurately.

Copied to clipboard

Xuan Chen, Hao Liu, Tao Yuan, Mehran Kafai, Piotr Habas, Xiangyu Zhang

AdaTooler-V: Adaptive Tool-Use for Images and Videos (2026.findings-acl)

Challenge:	Traditional phishing website detection relies on static heuristics or reference lists, which lag behind rapidly evolving attacks.
Approach:	They propose a memory-augmented multi-modal LLM agent that leverages episodic memories to guide decisions on recurring and novel threats.
Outcome:	The proposed agent outperforms state-of-the-art phishing detection tools on two public datasets and improves recall by 20%.

Copied to clipboard

Chaoyang Wang, Kaituo Feng, Dongyang Chen, Zhongyu Wang, Zhixun Li, Sicheng Gao, Meng Meng, Xu Zhou, Manyuan Zhang, Yuzhang Shang, Xiangyu Yue

Sanitizing Large Language Models in Bug Detection with Data-Flow (2024.findings-emnlp)

Challenge:	Existing models exhibit blind tool-use reasoning patterns, which significantly increases inference overhead and degrades model performance.
Approach:	They propose an MLLM that performs adaptive tool-use by determining whether a visual problem truly requires tools.
Outcome:	The proposed model outperforms existing methods in visual reasoning tasks.

Copied to clipboard

Chengpeng Wang, Wuqi Zhang, Zian Su, Xiangzhe Xu, Xiangyu Zhang

GeoArena: Evaluating Open-World Geographic Reasoning in Large Vision-Language Models (2026.acl-long)

Challenge:	Large language models (LLMs) have potential in code reasoning tasks but the hallucination effect can compromise the reliability of bug reports.
Approach:	They propose a new schema of bug detection that enforces LLMs to emit data-flow paths in few-shot chain-of-thought prompting and validates them via the program-property decomposition.
Outcome:	The proposed approach achieves 91.03% precision and 74.00% recall upon synthetic benchmarks and boosts precision by 21.99% with the sanitization.

Copied to clipboard

Pengyue Jia, Yingyi Zhang, Xiangyu Zhao, Sharon Li

An Evaluation Resource for Grounding Translation Errors (2025.findings-emnlp)

Challenge:	Existing evaluation paradigms for geographic reasoning are outcome-centric and focus on label matching, leaving the underlying linguistic reasoning chains as unexamined black boxes.
Approach:	They propose a dynamic, human-preference-based evaluation framework for benchmarking open-world geographic reasoning.
Outcome:	The proposed framework reframes evaluation as a pairwise reasoning alignment task on in-the-wild images, where human judges compare model-generated explanations based on reasoning quality, evidence synthesis, and plausibility.

Copied to clipboard

Sujin Chen, Kang Wang, Zixuan Zhou, Xiangyu Duan, Wanqun Zhang, Hao Yang, Jinsong Su, Min Zhang

Explainability & Interpretability in NLP

Challenge:	Current fine-grained error analyses do not ground the errors to the reasons why the annotated text spans are erroneous.
Approach:	They use a bi-directional grounding scheme to ground erroneous text in two directions . if the error spans of both directions are consistent, the explanation is valid .
Outcome:	The proposed grounding process improves translation error detection significantly.

LLMs Trust Humans More, That’s a Problem! Unveiling and Mitigating the Authority Bias in Retrieval-Augmented Generation (2025.acl-long)

Copied to clipboard

Yuxuan Li, Xinwei Guo, Jiashi Gao, Guanhua Chen, Xiangyu Zhao, Jiaxin Zhang, Quanying Liu, Haiyan Wu, Xin Yao, Xuetao Wei

Information Retrieval

Challenge:	Large language models (LLMs) generate outputs that stray from user input or contravene established knowledge.
Approach:	They propose a new phenomenon, Authority Bias, where LLMs favor one knowledge source over the other . they propose atomic information that generates conflicts and a Conflict Detection Enhanced Query framework .
Outcome:	The proposed framework reduces Authority bias in large language models . it detects conflicts, performs credibility assessment on conflicting paragraphs, and detects perturbed text .

Model Merging for Knowledge Editing (2025.acl-industry)

Copied to clipboard

Zichuan Fu, Xian Wu, Guojing Li, Yingying Zhang, Yefeng Zheng, Tianshi Ming, Yejing Wang, Wanyu Wang, Xiangyu Zhao

LLMTreeRec: Unleashing the Power of Large Language Models for Cold-Start Recommendations (2025.coling-main)

Challenge:	Existing knowledge editing approaches struggle with sequential editing scenarios and harm the general capabilities of the model.
Approach:	They propose a framework that combines robust supervised fine-tuning and model merging for knowledge editing to combine supervised and supervised learning.
Outcome:	The proposed approach outperforms existing methods in sequential editing while preserving the original performance of the model.

Copied to clipboard

Wenlin Zhang, Chuhan Wu, Xiangyang Li, Yuhao Wang, Kuicai Dong, Yichao Wang, Xinyi Dai, Xiangyu Zhao, Huifeng Guo, Ruiming Tang

Bilingual Dictionary Based Neural Machine Translation without Using Parallel Sentences (2020.acl-main)

Challenge:	Lack of training data leads to the system cold-start problem in recommendation systems, making them struggle to provide effective recommendations.
Approach:	They propose a tree-based LLM recommendation framework which structures all items into an item tree to improve the efficiency of LLM’s item retrieval.
Outcome:	The proposed framework outperforms the baseline model in the A/B test on Huawei industrial system.

Copied to clipboard

Xiangyu Duan, Baijun Ji, Hao Jia, Min Tan, Min Zhang, Boxing Chen, Weihua Luo, Yue Zhang

Multilinguality

Challenge:	a monolingual speaker can learn to translate by looking up a bilingual dictionary . a novel task of machine translation (MT) is based on no parallel sentences but can refer to a ground-truth bilingual dictionary and large-scale monolingual corpora.
Approach:	They propose a task of machine translation that uses a bilingual dictionary and large-scale monolingual corpora to translate a monolingual speaker.
Outcome:	The proposed task is based on a bilingual dictionary and large scale monolingual corpora, while being independent on parallel sentences.

Counterfactual Adversarial Learning with Representation Interpolation (2021.findings-emnlp)

Copied to clipboard

Wei Wang, Boxin Wang, Ning Shi, Jinfeng Li, Bingyu Zhu, Xiangyu Liu, Rong Zhang

Robustness in NLP

Challenge:	Existing models with statistical bias are prone to memorized correlations . large pre-trained models such as BERT have revolutionized the model development paradigm in natural language processing .
Approach:	They propose a framework to tackle the problem from a causal perspective using a latent space interpolation approach.
Outcome:	Extensive experiments show that CAT achieves substantial performance improvement over SOTA across different downstream tasks, including sentence classification, natural language inference and question answering.

PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning (2026.acl-long)

Copied to clipboard

Jingcheng Hu, Yinmin Zhang, Shijie Shang, Xiaobo Yang, Yue Peng, Zhewei Huang, Hebin Zhou, Xin Wu, Jie Cheng, Fanqi Wan, Xiangwen Kong, Chengyuan Yao, Kaiwen Yan, Ailin Huang, Hongyu Zhou, Qi Han, Zheng Ge, Xiangyu Zhang, Heung-Yeung Shum

Don’t Stop Early: Scalable Enterprise Deep Research with Controlled Information Flow and Evidence-Aware Termination (2026.acl-industry)

Challenge:	Parallel Coordinated Reasoning (PaCoRe) overcomes a central limitation of contemporary language models: their inability to scale test-time compute (TTC) far beyond sequential reasoning under a fixed context window.
Approach:	They propose a training-and-inference framework to overcome a central limitation of language models: their inability to scale test-time compute (TTC) under a fixed context window.
Outcome:	The proposed model scales to multi-million-token effective TTC without exceeding context limits.

Copied to clipboard

Prafulla Kumar Choubey, Kung-Hsiang Huang, Pranav Narayanan Venkit, Jiaxin Zhang, Vaibhav Vats, Yu Li, Xiangyu Peng, Chien-Sheng Wu

Challenge:	Enterprise deep research systems fail to produce decision-ready reports due to uneven information coverage, context explosion, and premature stopping.
Approach:	They propose a scalable Enterprise Deep Research (EDR) architecture that decomposes requests into coverage-driven objectives via outline generation with reflection and localizes context with dependency-guided execution and explicit information sharing.
Outcome:	The proposed system achieves the strongest overall performance compared with competitive deep-research baselines on internal sales enablement tasks and the public DeepResearch Bench benchmark.

Zero-Shot Cross-Lingual Abstractive Sentence Summarization through Teaching Generation and Attention (P19-1)

Copied to clipboard

Xiangyu Duan, Mingming Yin, Min Zhang, Boxing Chen, Weihua Luo

Beyond Sequences: Two-dimensional Representation and Dependency Encoding for Code Generation (2025.acl-long)

Challenge:	Abstractive Sentence Summarization (ASSUM) is a monolingual task that focuses on grasping the core idea of the source sentence and presenting it as the summary.
Approach:	They propose to use monolingual ASSUM to train a cross-lingual ASL system . they propose to train the system on summary word generation and attention .
Outcome:	Experiments show that the proposed method improves on the monolingual ASSUM task.

Copied to clipboard

Xiangyu Zhang, Yu Zhou, Guang Yang, Wei Cheng, Taolue Chen

Programming Languages in NLP

Challenge:	Existing code generation approaches represent code as a linear sequence of tokens, but positional encodings compromise generalization . explicit positional encoders sacrifice permutation invariance, imposes a strict order on the input sequence .
Approach:	They propose to represent code snippets as two-dimensional entities with explicit encodings . they propose to use dictionary learning to perform semantic matching between code lines .
Outcome:	The proposed model captures the hierarchical and spatial structure of code, especially the dependencies between code lines.

Mitigating Hallucinations of Large Language Models in Medical Information Extraction via Contrastive Decoding (2024.findings-emnlp)

Copied to clipboard

Derong Xu, Ziheng Zhang, Zhihong Zhu, Zhenxi Lin, Qidong Liu, Xian Wu, Tong Xu, Xiangyu Zhao, Yefeng Zheng, Enhong Chen

Information Extraction & Text Mining

Challenge:	Medical Information Extraction (MIE) tasks are a fundamental component of medical NLP.
Approach:	They propose an alternative adaptive constraint strategy to adjust the scale and scope of contrastive tokens.
Outcome:	The proposed approach selectively enhances the identification and classification capabilities while minimizing the influence of other inherent abilities in LLMs.

Token Drop mechanism for Neural Machine Translation (2020.coling-main)

Copied to clipboard

Huaao Zhang, Shigui Qiu, Xiangyu Duan, Min Zhang

Machine Translation

Challenge:	Neural machine translation models are vulnerable to unfamiliar inputs.
Approach:	They propose to drop tokens of the input sentences to improve generalization and avoid overfitting for the NMT model.
Outcome:	The proposed approach improves on Chinese-English and English-Romanian benchmarks and achieves significant performance improvements over baselines.

From AR to Diffusion: Efficiently Adapting Large Language Models with Strictly Causal and Elastic Horizons (2026.acl-long)

Copied to clipboard

Xiangyu Ma, Teng Xiao, Zuchao Li, Lefei Zhang

Threat Behavior Textual Search by Attention Graph Isomorphism (2024.eacl-long)

Challenge:	Autoregressive (AR) models rely on bidirectional attention, creating a structural mismatch with pre-trained Autoregression models.
Approach:	They propose a framework that efficiently adapts autoregressive (AR) models to the diffusion paradigm.
Outcome:	The proposed framework reduces training costs by orders of magnitude while maintaining state-of-the-art performance.

Copied to clipboard

Chanwoo Bae, Guanhong Tao, Zhuo Zhang, Xiangyu Zhang

Multi-matrix Factorization Attention (2025.findings-acl)

Challenge:	Existing methods to analyze malware behavior only disclose a subset of behaviors due to inherent difficulties.
Approach:	They propose a novel malware behavior search technique that is based on graph isomorphism at the attention layers of Transformer models.
Outcome:	The proposed technique outperforms state-of-the-art methods in a case study of 10 real-world malwares by 6-14%.

Copied to clipboard

Jingcheng Hu, Houyi Li, Yinmin Zhang, Zili Wang, Shuigeng Zhou, Xiangyu Zhang, Heung-Yeung Shum

Challenge:	Existing variants for Multi-Head Attention (MHA) fail to maintain strong performance under stringent Key-Value cache (KV cache) constraints.
Approach:	They propose to use multi-matrix factorization attention and MFA-Key-reuse attention architectures to increase model capacity under tight KV cache constraints.
Outcome:	The proposed architecture outperforms existing methods while reducing KV cache usage by 56% and 93.7% in large-scale experiments.

System Prompt Hijacking via Permutation Triggers in LLM Supply Chains (2025.findings-acl)

Copied to clipboard

Lu Yan, Siyuan Cheng, Xuan Chen, Kaiyuan Zhang, Guangyu Shen, Xiangyu Zhang

ThinkAnswer Loss: Balancing Semantic Similarity and Exact Matching for LLM Reasoning Enhancement (2025.findings-emnlp)

Challenge:	Large language models (LLMs) are rapidly transforming the landscape of artificial intelligence due to the substantial resources required for training.
Approach:	They propose a post-deployment attack that bypasses system prompts to compromise models . they introduce Precise Activation Guarding and Unit Deviation Sampling to protect against attack .
Outcome:	The proposed attack bypasses system prompts, enabling unrestricted model outputs and safety violations.

Copied to clipboard

Shan Yang, Kun Wu, Zeju Li, Linlin Zhang, Xiangyu Pei, Leike An, Yu Liu

When LLMs Meets Acoustic Landmarks: An Efficient Approach to Integrate Speech into Large Language Models for Depression Detection (2024.emnlp-main)

Challenge:	Existing methods for knowledge distillation use Chain-of-Thought (CoT) and answer pairs, but they lack appropriate supervision signals.
Approach:	They propose a framework that decouples CoT and answer supervision . the framework applies semantic similarity constraints while maintaining strict literal matching for the answer .
Outcome:	The proposed framework decouples CoT and answer supervision while maintaining strict literal matching for the answer.

Copied to clipboard

Xiangyu Zhang, Hexin Liu, Kaishuai Xu, Qiquan Zhang, Daijiao Liu, Beena Ahmed, Julien Epps

Backdooring Neural Code Search (2023.acl-long)

Challenge:	Large language models (LLMs) are used for depression detection but their application remains unexplored.
Approach:	They propose to integrate acoustic speech information into LLMs for depression detection by integrating aural landmarks into the framework.
Outcome:	The proposed method adds critical dimensions to speech transcripts and provides insights into the unique speech patterns of individuals.

Copied to clipboard

Weisong Sun, Yuchen Chen, Guanhong Tao, Chunrong Fang, Xiangyu Zhang, Quanjun Zhang, Bin Luo

Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory (2026.acl-long)

Challenge:	Neural code search models are used to find code snippets from online repositories . however, their security aspect is rarely studied .
Approach:	They propose to use off-the-shelf code snippets from online repositories to find desired code . they propose to inject a backdoor into neural code search models which return buggy code if attacker modifies one variable/function name .
Outcome:	The proposed attack outperforms baselines on two neural code search models by 60%.

Copied to clipboard

Derong Xu, Shuochen Liu, Pengfei Luo, Pengyue Jia, Yingyi Zhang, Yi Wen, Yimin Deng, Wenlin Zhang, Enhong Chen, Xiangyu Zhao, Tong Xu

Challenge:	Existing memory systems rely on static, hand-crafted update rules for personalization, but sparse outcome rewards provide weak supervision, resulting in unstable long-horizon optimization.
Approach:	They propose a memory guideline optimization framework that learns how memory should be organized and what information to update.
Outcome:	The proposed framework learns how memory should be organized and what information to update.

Multimodal Cross-lingual Phrase Retrieval (2024.lrec-main)

Copied to clipboard

Chuanqi Dong, Wenjie Zhou, Xiangyu Duan, Yuqi Zhang, Min Zhang

Improving Embedding-based Large-scale Retrieval via Label Enhancement (2021.findings-emnlp)

Challenge:	Existing approaches to cross-lingual phrase retrieval only deal with textual modality, leaving the question of the effectiveness of using multimodal information unanswered.
Approach:	They propose a multimodal cross-lingual phrase retrieval resource that integrates a Wikimedia Commons media store and a large multimodal pre-trained model to bridge the gap between different modalities.
Outcome:	The proposed approach performs significantly better than pure textual cross-lingual phrase retrieval on a benchmarked dataset covering eight language pairs.

Copied to clipboard

Peiyang Liu, Xi Wang, Sen Wang, Wei Ye, Xiangyu Xi, Shikun Zhang

Representation Learning

Challenge:	Existing methods for large-scale retrieval are trained with 0-1 hard labels that indicate whether a query is relevant to a document, ignoring rich information of the relevance degree.
Approach:	They propose to introduce label enhancement for the first time to characterize query-document relevance degree by embedding label distribution into contextual embeddables.
Outcome:	The proposed method significantly outperforms existing retrieval models and its counterparts equipped with two alternative methods on English and Chinese large-scale retrieval tasks.

Can LLMs Hear the Dogwhistle? (2026.findings-acl)

Copied to clipboard

Yifan Liu, Yi Lin, Xinwei Guo, Ziwei Wang, Jiaxin Zhang, Guanhua Chen, Haiyan Wu, Xiangyu Zhao, Xin Yao, Xuetao Wei

UniFashion: A Unified Vision-Language Model for Multimodal Fashion Retrieval and Generation (2024.emnlp-main)

Challenge:	Existing safety benchmarks focus on explicitly harmful content, but ignore context-dependent expressions such as dogwhistles.
Approach:	They propose a benchmark for evaluating LLM safety under dogwhistle-driven prompts . their findings expose a blind spot in current safety evaluation practices .
Outcome:	The proposed benchmark compared safety performance with toxic terms using dogwhistle-driven prompts.

Copied to clipboard

Xiangyu Zhao, Yuehan Zhang, Wenlong Zhang, Xiao-Ming Wu

MemSearch-o1: Empowering Large Language Models with Reasoning-Aligned Memory Growth in Agentic Search (2026.acl-long)

Challenge:	e-commerce tasks such as multimodal retrieval and multimodal generation are largely ignored due to the diversity of the multimodal fashion domain.
Approach:	They propose a framework that integrates image generation with retrieval and text generation tasks.
Outcome:	The proposed framework outperforms state-of-the-art models across fashion tasks.

Copied to clipboard

Sheng Zhang, Junyi Li, Yingyi Zhang, Pengyue Jia, Yichao Wang, Xiaowei Qian, Wenlin Zhang, Maolin Wang, Yong Liu, Xiangyu Zhao

Third-Party Aligner for Neural Word Alignments (2022.findings-emnlp)

Challenge:	Existing methods for memory management struggle to capture fine-grained semantic relations between queries and documents.
Approach:	They propose a framework for reasoning and agentic search that grows fine-grained memory fragments from seed tokens from queries, then retraces and deep refines the memory via a contribution function.
Outcome:	Experiments on eight benchmark datasets show that MemSearch-o1 significantly mitigates memory dilution and more effectively activates reasoning potential of diverse LLMs.

Copied to clipboard

Jinpeng Zhang, Chuanqi Dong, Xiangyu Duan, Yuqi Zhang, Min Zhang

Multilinguality

Challenge:	Existing work shows that word alignment can be competitive .
Approach:	They propose to use word alignments generated by a third-party word aligner to supervise the neural word alignment training.
Outcome:	The proposed approach can find more accurate word alignments and delete wrong alignments, leading to better performance than the current best third-party word aligner.

Syntax-Aware Retrieval Augmented Code Generation (2023.findings-emnlp)

Copied to clipboard

Xiangyu Zhang, Yu Zhou, Guang Yang, Taolue Chen

Programming Languages in NLP

Challenge:	Neural code generation models with token-level retrieval capabilities are often noisy and time-consuming.
Approach:	They propose a token-level retrieval augmented code generation method that leverages syntax constraints for the retrieval of datastores.
Outcome:	The proposed method reduces the impact of retrieve noise on code generation on two datasets.

PRIME: A Process-Outcome Alignment Benchmark for Verifiable Reasoning in Mathematics and Engineering (2026.acl-long)

Copied to clipboard

Xiangfeng Wang, Hangyu Guo, Yanlin Lai, Mitt Huang, Liang Zhao, Chengyuan Yao, Yinmin Zhang, Qi Han, null Xiaoxiaoren, Chun Yuan, Tong Xu, Zheng Ge, Xiangyu Zhang, Daxin Jiang

HSGraphAgent: Knowledge-Graph-Guided Large Language Models for Harmonized System Code Classification (2026.acl-long)

Challenge:	Current outcome-centric verification paradigms neglect potential errors in the derivation process.
Approach:	They propose a process-aware RLVR training paradigm utilizing verifiers selected via PRIME.
Outcome:	The proposed approach outperforms the baseline verification paradigm on AIME24, AIME25, and Beyond-AIME models.

Copied to clipboard

Qiang Xia, Zijian Zhang, Ao Wang, Wenhan Wang, Xiangyu Wang, Jian Li

Graph Enhanced Dual Attention Network for Document-Level Relation Extraction (2020.coling-main)

Challenge:	Harmonized System (HS) code classification is a hierarchically structured and regulation-constrained task, often complicated by short and noisy product descriptions.
Approach:	They propose a knowledge-graph-guided LLM framework that formulates HS classification as a stepwise, regulation-aware reasoning process over an explicit HS knowledge graph.
Outcome:	The proposed framework constrains inference to legally valid paths while producing explicit and traceable reasoning trajectories.

Copied to clipboard

Bo Li, Wei Ye, Zhonghao Sheng, Rui Xie, Xiangyu Xi, Shikun Zhang

Information Extraction & Text Mining

Challenge:	Document-level relation extraction requires inter-sentence reasoning capabilities to capture local and global contextual information for multiple relation facts.
Approach:	They propose to characterize the interaction between sentences and potential relation instances via a Graph Enhanced Dual Attention network (GEDA) . they also propose a simple yet effective regularizer based on the natural duality of the S2R and R2S attentions, whose weights are also supervised by the supporting evidence of relation instances during training.
Outcome:	The proposed model achieves competitive performance on an existing large-scale dataset while the predictions can be interpretable and easily observed.

RU22Fact: Optimizing Evidence for Multilingual Explainable Fact-Checking on Russia-Ukraine Conflict (2024.lrec-main)

Copied to clipboard

Yirong Zeng, Xiao Ding, Yi Zhao, Xiangyu Li, Jie Zhang, Chao Yao, Ting Liu, Bing Qin

MedDialog: Large-scale Medical Dialogue Datasets (2020.emnlp-main)

Challenge:	Existing methods to verify factuality of claims do not provide sufficient evidence for explainable fact-checking systems.
Approach:	They propose a method to automatically retrieve and summarize evidence from the Web and a novel multilingual explainable fact-checking dataset on the Russia-Ukraine conflict in 2022.
Outcome:	The proposed method can retrieve and summarize evidence from the Web and generate explanations in 16 languages.

Copied to clipboard

Guangtao Zeng, Wenmian Yang, Zeqian Ju, Yue Yang, Sicheng Wang, Ruisi Zhang, Meng Zhou, Jiaqi Zeng, Xiangyu Dong, Ruoyu Zhang, Hongchao Fang, Penghui Zhu, Shu Chen, Pengtao Xie

Dialogue Systems & Conversational Agents

Challenge:	telemedicine is a medical practice that provides patient care remotely using video conferencing tools.
Approach:	They build large-scale medical dialogue datasets to facilitate research . they pretrain several models on the Chinese MedDialog dataset and compare their performance .
Outcome:	The proposed datasets show that models trained on MedDialog can generate doctor-like medical dialogues.

Logic-Consistency Text Generation from Semantic Parses (2021.findings-acl)

Copied to clipboard

Chang Shu, Yusen Zhang, Xiangyu Dong, Peng Shi, Tao Yu, Rui Zhang

Text Generation

Challenge:	Text generation from semantic parses is challenging due to the complexity of the inner logic and the lack of automatic evaluation metrics for logic consistency.
Approach:	They propose a framework for logic consistent text generation from semantic parses that employs iterative training procedures and quality control.
Outcome:	The proposed framework enhances logic consistency and human evaluation on two benchmark datasets.

Combining Static Word Embeddings and Contextual Representations for Bilingual Lexicon Induction (2021.findings-acl)

Copied to clipboard

Jinpeng Zhang, Baijun Ji, Nini Xiao, Xiangyu Duan, Min Zhang, Yangbin Shi, Weihua Luo

Multilinguality

Challenge:	Bilingual Lexicon Induction (BLI) aims to map words in one language to their translations in another.
Approach:	They propose a mechanism to combine static word embeddings and contextual representations to utilize the advantages of both paradigms.
Outcome:	The proposed method improves performance on supervised and unsupervised BLI benchmarks on all language pairs by average improving 3.2 points over baselines.

Raw Pointer Rewriting with LLMs for Translating C to Safer Rust (2026.findings-acl)

Copied to clipboard

Yifei Gao, Chengpeng Wang, Pengxiang Huang, Xuwei Liu, Mingwei Zheng, Xiangyu Zhang

Text Generation

Challenge:	C2Rust is a system programming language that enforces strict memory and type safety guarantees.
Approach:	They propose a raw pointer rewriting technique that lifts raw pointers in individual functions to appropriate Rust data structures.
Outcome:	The proposed technique eliminates 18.57% of local raw pointers and improves memory safety on 28 real-world C projects.

Beyond Overlap Metrics: Rewarding Reasoning and Preferences for Faithful Multi-Role Dialogue Summarization (2026.findings-acl)

Copied to clipboard

Xiaoyong Mei, Tingting Zuo, Da Chen, Guangyu Hu, Xiangyu Wen, Chao Duan, Mingyan Zhang, Fudan Zheng

Dialogue Systems & Conversational Agents

Challenge:	Existing methods for multi-role dialogue summarization favor surface-level imitation of references rather than genuine gains in faithfulness or alignment with human preferences.
Approach:	They propose a framework that couples explicit cognitive-style reasoning with reward-based optimization for multi-role dialogue summarization.
Outcome:	The proposed framework matches strong baselines on ROUGE and BERTScore, while in-depth analysis on SAMSum shows clear gains in factual faithfulness and model-based preference alignment.

Label Smoothing for Text Mining (2022.coling-1)

Copied to clipboard

Peiyang Liu, Xiangyu Xi, Wei Ye, Shikun Zhang

Information Retrieval

Challenge:	Existing text mining models are trained with 0-1 hard label that indicates whether an instance belongs to a class, ignoring rich information of the relevance degree.
Approach:	They propose a keyword-based method to automatically generate soft labels from hard labels . they exploit relevance between labels and instances to incorporate them into models .
Outcome:	The proposed method improves models under balanced and unbalanced conditions.

Submodular-based In-context Example Selection for LLMs-based Machine Translation (2024.lrec-main)

Copied to clipboard

Baijun Ji, Xiangyu Duan, Zhenyu Qiu, Tong Zhang, Junhui Li, Hao Yang, Min Zhang

Machine Translation

Challenge:	Prior studies have focused on the role of well-chosen examples in in-context learning .
Approach:	They propose to use multiple translational factors for in-context example selection by using monotone submodular function maximization.
Outcome:	The proposed approach outperforms random selection and robust single-factor baselines across various NLP tasks.

WISCA: A Lightweight Model Transition Method to Improve LLM Training via Weight Scaling (2026.findings-acl)

Copied to clipboard

Jiacheng Li, Jianchao Tan, Zhidong Yang, Pingwei Sun, Feiye Huo, Jiayu Qin, Xiangyu Zhang, Maoxin He, Guangming Tan, Weile Jia, Xunliang Cai, Tong Zhao

RoChBert: Towards Robust BERT Fine-tuning for Chinese (2022.findings-emnlp)

Challenge:	Recent advances in training optimization for Transformer-based large language models lack systematic optimization of weight patterns during training.
Approach:	They propose a Weight Scaling method that rescales weights while preserving model outputs to improve model training efficiency and model quality.
Outcome:	The proposed method significantly improves convergence quality and loss reduction in LLMs with Grouped Query Attention architectures and LoRA fine-tuning tasks.

Copied to clipboard

Zihan Zhang, Jinfeng Li, Ning Shi, Bo Yuan, Xiangyu Liu, Rong Zhang, Hui Xue, Donghong Sun, Chao Zhang

Exploring Reasoning Reward Model for Agents (2026.findings-acl)

Challenge:	Pre-trained language models (e.g., BERT) have been proved vulnerable to adversarial texts.
Approach:	They propose to fuse Chinese phonetic and glyph features into pre-trained models by using a more comprehensive adversarial graph.
Outcome:	The proposed framework outperforms existing methods in significant ways on a wide range of tasks while remaining accurate on benign texts.

Copied to clipboard

Kaixuan Fan, Kaituo Feng, Manyuan Zhang, Tianshuo Peng, Zhixun Li, Yilei Jiang, Shuang Chen, Xiangyu Yue

BalanceSFT: Improving LLM Function Calling with Balanced Training Signals and Data Hardness (2026.findings-acl)

Challenge:	Existing methods for agentic reinforcement learning rely on sparse outcome-based reward for training, leading to suboptimal results.
Approach:	They propose an agent-based reward model that produces structured feedback for agentic trajectories, including an explicit reasoning trace and a focused critique.
Outcome:	The proposed model produces structured feedback for agentic trajectories including an explicit reasoning trace, a focused critique, and an overall score that evaluates process performance.

Copied to clipboard

Bingguang Hao, Zengzhuang Xu, Maolin Wang, Yuntao Wen, Yicheng Chen, Cunyin Peng, Long Chen, Xiangyu Zhao, Jinjie Gu, Chenyi Zhuang, Ji Zhang

Redundancy Principles for MLLMs Benchmarks (2025.acl-long)

Challenge:	Currently, Supervised Fine-Tuning (SFT) is the prevailing method for equipping Large Language Models (LLMs) with function calling capabilities, but its effectiveness is often compromised by two challenges: 1) lengthy Chain-of-Thought (CoT) reasoning tokens dominate training signals over concise function calls in the learning objective; 2) scarcity of hard training examples.
Approach:	They propose a framework that uses a self-adjusted signal balancing loss and a hard data re-sampling strategy to selectively generate new, high-quality complex data guided by model errors.
Outcome:	The proposed framework surpasses state-of-the-art models like GPT-5 in function calling performance.

Copied to clipboard

Zicheng Zhang, Xiangyu Zhao, Xinyu Fang, Chunyi Li, Xiaohong Liu, Xiongkuo Min, Haodong Duan, Kai Chen, Guangtao Zhai

Basic Reading Distillation (2025.acl-long)

Challenge:	Rapid growth of Multi-modality Large Language Models has led to significant redundancy among benchmarks.
Approach:	They propose a framework to improve MLLM benchmark design by identifying redundancy at three levels: dimension, instance, and cross-benchmark redundancies.
Outcome:	The proposed framework streamlines evaluations and enhances reliability.

Copied to clipboard

Zhi Zhou, Sirui Miao, Xiangyu Duan, Hao Yang, Min Zhang

Contrastive Attention Mechanism for Abstractive Sentence Summarization (D19-1)

Challenge:	Large language models require high computational resources which limits their deployment in real-world applications.
Approach:	They propose to distill large language models into smaller language models by either knowledge distillation or task distillation.
Outcome:	The proposed model outperforms or performs comparable to over 20x bigger LLMs on language inference benchmarks and BIG-bench tasks.

Copied to clipboard

Xiangyu Duan, Hongfei Yu, Mingming Yin, Min Zhang, Weihua Luo, Yue Zhang

Summarization

Challenge:	Existing attention mechanisms for abstractive sentence summarization are based on rule-based methods and large-scale training corpora.
Approach:	They propose a contrastive attention mechanism that extends the sequence-to-sequence framework for abstractive sentence summarization task.
Outcome:	The proposed mechanism improves the state-of-the-art on the abstractive sentence summarization task.

T2R-BENCH: A Benchmark for Real World Table-to-Report Task (2025.emnlp-main)

Copied to clipboard

Jie Zhang, Changzai Pan, Sishi Xiong, Kaiwen Wei, Yu Zhao, Xiangyu Li, Jiaxin Peng, Xiaoyan Gu, Jian Yang, Wenhan Chang, Zhenhe Wu, Jiang Zhong, Shuangyong Song, Xuelong Li

Structured Data in NLP

Challenge:	Existing table benchmarks lack the capacity to adequately assess the practical application of table reasoning in industrial applications.
Approach:	They propose a bilingual table-to-report task and a table-based benchmark to assess the quality of table reasoning.
Outcome:	The proposed task is based on a bilingual benchmark with 457 industrial tables and evaluation criteria to measure the quality of report generation.

SpeechT-RAG: Reliable Depression Detection in LLMs with Retrieval-Augmented Generation Using Speech Timing Information (2025.findings-acl)

Copied to clipboard

Xiangyu Zhang, Hexin Liu, Qiquan Zhang, Beena Ahmed, Julien Epps

MirrorCAPTCHA: Wild CAPTCHA, Wild Distribution, Wild Web-based Platform Meet Multimodal LLM Agents (2026.acl-long)

Challenge:	Large Language Models (LLMs) have been extensively utilized for health-related tasks, yet their performance in depression detection remains limited when relying solely on text input.
Approach:	They propose a system that leverages speech timing features for depression detection and reliable confidence estimation.
Outcome:	The proposed system outperforms text-based RAG systems in depression detection and confidence estimation.

Copied to clipboard

Xiangyu Wu, Yuwei Hu, Tianyu Cui, Yueying Tian, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang, Yang Yang, Jianfeng Lu

Unlocking Exploration in RLVR: Uncertainty-aware Advantage Shaping for Deeper Reasoning (2026.findings-acl)

Challenge:	Existing agent benchmarks fail to evaluate an agent's real-world capacity to handle CAPTCHA . Existing benchmarks ignore this practical challenge, failing to evaluate agents' ability to handle complex visual CAPTchas.
Approach:	They propose a benchmark annotated with Weighted Pass Rate and a new metric to measure agent's ability to handle CAPTCHA.
Outcome:	The proposed benchmark outperforms current state-of-the-art closed-source models on mirrorCAPTCHA and achieves 9.4% higher average weighted pass rate and 2.13% higher average Completion degree.

Copied to clipboard

Can Xie, Ruotong Pan, Xiangyu Wu, Zhang Yunfei, Jiayi Fu, Tingting Gao, Guorui Zhou

Disambiguated Lexically Constrained Neural Machine Translation (2023.findings-acl)

Challenge:	Reinforcement Learning with Verifiable Rewards (RLVR) has shown significant promise for enhancing the reasoning capabilities of large language models (LLMs).
Approach:	They propose a model-free method that refines credit assignment by leveraging the model's internal uncertainty signals.
Outcome:	Extensive experiments on five mathematical reasoning benchmarks show that the proposed method outperforms strong RLVR baselines on multiple model scales, including 1.5B and 7B.

Copied to clipboard

Jinpeng Zhang, Nini Xiao, Ke Wang, Chuanqi Dong, Xiangyu Duan, Yuqi Zhang, Min Zhang

Machine Translation

Challenge:	Current approaches to LCNMT assume that pre-specified lexicon constraints are contextually appropriate.
Approach:	They propose a framework that disambiguates constraints based on contexts at first and integrates them into LCNMT.
Outcome:	The proposed approach outperforms baseline approaches on benchmark datasets and comprehensive experiments in multiple target constraints.

Efficient Dynamic Clustering-Based Document Compression for Retrieval-Augmented-Generation (2025.findings-emnlp)

Copied to clipboard

Weitao Li, Xiangyu Zhang, Kaiming Liu, Xuanyu Lei, Weizhi Ma, Yang Liu

Green & Sustainable NLP

Challenge:	RAG implementations face challenges in addressing retrieved noise and redundant content . current RAG methods lack the ability to exploit fine-grained inter-document relationships .
Approach:	They propose a retrieval-augmented generation framework that exploits latent inter-document relationships while removing irrelevant information and redundant content.
Outcome:	The proposed framework achieves consistent performance improvements on knowledge-QA and hallucination-Detection datasets.

Speaking in Wavelet Domain: A Simple and Efficient Approach to Speed up Speech Diffusion Model (2024.emnlp-main)

Copied to clipboard

Xiangyu Zhang, Daijiao Liu, Hexin Liu, Qiquan Zhang, Hanyu Meng, Leibny Paola Garcia Perera, EngSiong Chng, Lina Yao

Stepwise Reasoning Disruption Attack of LLMs (2025.acl-long)

Challenge:	Existing approaches to enhance inference speed and training require complex modifications to the model.
Approach:	They propose to double the training and inference speed of Denoising Diffusion Probabilistic Models by simply redirecting the generative target to the wavelet domain.
Outcome:	The proposed method doubles the training and inference speed of Speech DDPMs by redirecting the generative target to the wavelet domain.

Copied to clipboard

Jingyu Peng, Maolin Wang, Xiangyu Zhao, Kai Zhang, Wanyu Wang, Pengyue Jia, Qidong Liu, Ruocheng Guo, Qi Liu

Foot-In-The-Door: A Multi-turn Jailbreak for LLMs (2025.emnlp-main)

Challenge:	Existing attacks on LLM reasoning are constrained by specific settings or lack of imperceptibility, limiting their feasibility and generalizability.
Approach:	They propose a stepwise rEasoning error disruption attack that subtly injects errors into prior reasoning steps to mislead the model into producing incorrect subsequent reasoning and final answers.
Outcome:	The proposed attack is compatible with zero-shot and few-shot settings, maintains the natural reasoning flow, and ensures covert execution without modification of the instruction.

Copied to clipboard

Zixuan Weng, Xiaolong Jin, Jinyuan Jia, Xiangyu Zhang

OneRec-Think: In-Text Reasoning for Generative Recommendation (2026.acl-long)

Challenge:	Large Language Models (LLMs) are increasingly integrated into real-world applications, requiring a high level of safety and alignment.
Approach:	They propose a multi-turn jailbreak method that leverages foot-in-the-door principles to escalate malicious intent of user queries through intermediate bridge prompts and aligns the model’s response by itself to induce toxic responses.
Outcome:	The proposed method achieves an average attack success rate of 94% across seven widely used models outperforming existing state-of-the-art methods.

Copied to clipboard

Zhanyu Liu, Shiyao Wang, Xingmei Wang, Rongzhou Zhang, Jiaxin Deng, Honghui Bao, Jinghao Zhang, Wuchao Li, PengFei Zheng, Xiangyu Wu, Yifei Hu, Qigen Hu, Xinchen Luo, Lejian Ren, Zhang Zixing, Qianqian Wang, Kuo Cai, Yunfan Wu, Hongtao Cheng, Zexuan Cheng, Lu Ren, Huanjie Wang, Yi Su, Ruiming Tang, Kun Gai, Guorui Zhou

SampleMix: A Sample-wise Pre-training Data Mixing Strategy by Coordinating Data Quality and Diversity (2025.findings-emnlp)

Challenge:	Existing generative models lack the capacity for explicit and controllable reasoning, a key advantage of LLMs.
Approach:	They propose a framework that integrates dialogue, reasoning, and personalized recommendation.
Outcome:	Experiments across public benchmarks show state-of-the-art performance.

Copied to clipboard

Xiangyu Xi, Deyang Kong, Jian Yang, Jiawei Yang, Zhengyu Chen, Wei Wang, Jingang Wang, Xunliang Cai, Shikun Zhang, Wei Ye

AI4Reading: Chinese Audiobook Interpretation System Based on Multi-Agent Collaboration (2025.acl-demo)

Challenge:	Existing methods for pretraining data mixing for large language models neglect significant inter-domain overlaps and commonalities, failing to control the global diversity of the constructed training dataset.
Approach:	They propose a sample-wise data mixture approach that performs global cross-domain sampling by systematically evaluating the quality and diversity of each sample.
Outcome:	The proposed method exceeds existing domain-based methods in multiple downstream tasks and perplexity assessments.

Copied to clipboard

Minjiang Huang, Jipeng Qiang, Yi Zhu, Chaowei Zhang, Xiangyu Zhao, Kui Yu

Profiler: Black-box AI-generated Text Origin Detection via Context-aware Inference Pattern Analysis (2025.emnlp-main)

Challenge:	Interpretative audiobooks are becoming more popular, but their manual creation process remains time-consuming and resource-intensive.
Approach:	They propose a multi-agent collaboration system that leverages large language models and speech synthesis technology to generate podcast-like audiobook interpretations.
Outcome:	The proposed system is open source and open to the public.

Copied to clipboard

Hanxi Guo, Siyuan Cheng, Xiaolong Jin, Zhuo Zhang, Guangyu Shen, Kaiyuan Zhang, Shengwei An, Guanhong Tao, Xiangyu Zhang

Challenge:	Existing methods to identify the origin of AI-generated texts fail to identify origin due to the high similarity of different LLMs.
Approach:	They propose a black-box AI-generated text origin detection method which accurately predicts the origin of an input text by extracting distinct context inference patterns.
Outcome:	The proposed method outperforms 10 state-of-the-art baselines and achieves a 25% increase in AUC score on average across natural language and code datasets.

Exploiting the Shadows: Unveiling Privacy Leaks through Lower-Ranked Tokens in Large Language Models (2025.acl-long)

Copied to clipboard

Yuan Zhou, Zhuo Zhang, Xiangyu Zhang

Multi-perspective Improvement of Knowledge Graph Completion with Large Language Models (2024.lrec-main)

Challenge:	Large language models face vulnerabilities related to the extraction of sensitive information.
Approach:	They propose a method to exploit the model's lower-ranked output tokens to extract private information from retrieved documents or training knowledge.
Outcome:	The proposed method is effective in both the agentic application privacy extraction setting and the direct training data extraction.

Copied to clipboard

Derong Xu, Ziheng Zhang, Zhenxi Lin, Xian Wu, Zhihong Zhu, Tong Xu, Xiangyu Zhao, Yefeng Zheng, Enhong Chen

Logic Jailbreak: Efficiently Unlocking LLM Safety Restrictions Through Formal Logical Expression (2026.findings-acl)

Challenge:	Knowledge graph completion (KGC) is a widely used method to tackle incompleteness in knowledge graphs (KGs).
Approach:	They propose a general framework to compensate for the deficiency of contextualized knowledge by querying large language models from various perspectives.
Outcome:	The proposed framework improves knowledge graph completion (KGC) by querying large language models from various perspectives.

Copied to clipboard

Jingyu Peng, Maolin Wang, Nan Wang, Jiatong Li, Yuchen Li, Yuyang Ye, Wanyu Wang, Pengyue Jia, Kai Zhang, Xiangyu Zhao

OmniAlign-V: Towards Enhanced Alignment of MLLMs with Human Preference (2025.acl-long)

Challenge:	Despite advances in aligning LLMs with human values, current safety mechanisms remain vulnerable to jailbreak attacks.
Approach:	They propose a black-box jailbreak method that uses logical expression translation to bypass LLM safety mechanisms.
Outcome:	The proposed method exploits the distributional gap between alignment data and logic-expressed inputs while preserving the underlying semantic intent and readability while evading safety constraints.

Copied to clipboard

Xiangyu Zhao, Shengyuan Ding, Zicheng Zhang, Haian Huang, Maosongcao Maosongcao, Jiaqi Wang, Weiyun Wang, Xinyu Fang, Wenhai Wang, Guangtao Zhai, Hua Yang, Haodong Duan, Kai Chen