Papers by Zhang Mei
GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs (2026.acl-long)
Copied to clipboard
Weidong Tang, Jierui Li, Yueling Hou, Zihan Mei, Can Zhang, Xinyan Wan, Zhiyuan Liang, Pengfei Zhou, Yang You, Wangbo Zhao
| Challenge: | Existing models for general intelligence fail to model how mental states interact and crystallize into group-level outcomes. |
| Approach: | They propose a multimodal benchmark for group-level Theory of Mind (ToM) to probe nonlinear collective behavior. |
| Outcome: | The proposed model performs significantly below human levels, exposing blind spots in modeling social structures and nonlinear collective behavior. |
SMART: Evaluating LLMs’ Mathematical Reasoning via a Human Cognitive Process-Inspired Benchmark (2026.acl-long)
Copied to clipboard
| Challenge: | Existing evaluation methods focus on the final answer or on the intermediate reasoning steps, overlooking its inherently multi-stage and multi-dimensional nature. |
| Approach: | They propose a benchmark that decomposes mathematical problem-solving into four cognitive dimensions and introduces dimension-specific tasks to measure their cognitive processes. |
| Outcome: | The proposed model decomposes mathematical problem-solving into four cognitive dimensions and introduces dimension-specific tasks to measure their cognitive processes. |
NOTABLE: Transferable Backdoor Attacks Against Prompt-based NLP Models (2023.acl-long)
Copied to clipboard
| Challenge: | Existing backdoor attacks against prompt-based learning involve injecting back doors into embedding layers or word embedders. |
| Approach: | They propose a backdoor attack against prompt-based learning that injects backdoors into embedding layers or word embeddable vectors. |
| Outcome: | The proposed backdoor attack outperforms two state-of-the-art models on six NLP tasks and three prompting strategies. |
Context-DPO: Aligning Language Models for Context-Faithfulness (2025.findings-acl)
Copied to clipboard
Baolong Bi, Shaohan Huang, Yiwei Wang, Tianchi Yang, Zihan Zhang, Haizhen Huang, Lingrui Mei, Junfeng Fang, Zehao Li, Furu Wei, Weiwei Deng, Feng Sun, Qi Zhang, Shenghua Liu
| Challenge: | Context-DPO is the first alignment method specifically designed to enhance contextfaithfulness for large language models. |
| Approach: | They propose a benchmark that simulates Retrieval-Augmented Generation scenarios with knowledge conflicts to evaluate context-faithfulness. |
| Outcome: | The proposed method improves LLMs' context-faithfulness by 35% to 280% over open-source models. |
Reasoning about Uncertainty: Do Reasoning Models Know When They Don’t Know? (2026.findings-eacl)
Copied to clipboard
| Challenge: | Reasoning models are prone to generating confident, plausible responses that are incorrect (hallucinations). |
| Approach: | They introduce introspective uncertainty quantification to examine whether reasoning models are well-calibrated and does deeper reasoning improve their calibration? |
| Outcome: | The proposed model calibrations show that models are overconfident, overconfent and overconfust with deeper reasoning. |
ExAnte: A Benchmark for Ex-Ante Inference in Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) struggle with ex-ante reasoning—making inferences or predictions without access to future information. |
| Approach: | They propose a benchmark that assesses LLMs’ ex-ante inference ability across four tasks: stock prediction, question answering, Wikipedia event generation, and scientific publication generation. |
| Outcome: | The proposed benchmark assesses LLMs’ ex-ante inference ability across four tasks. |
Forget the Unneeded: Backdooring Large Language Models via Contrastive-enhanced Machine Unlearning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for prompt tuning for Large Language Models find backdoor attacks to be significant in data-rich scenarios. |
| Approach: | They propose a backdoor attacks through contrastive-enhanced machine unlearning in data-limited scenarios . they use a machine un learning method to capture precise backdoor patterns . |
| Outcome: | The proposed method captures precise backdoor patterns without association between triggers and backdoors, reducing side effects. |
ADEPT-SQL: A High-performance Text-to-SQL Application for Real-World Enterprise-Level Databases (2025.acl-demo)
Copied to clipboard
Yongnan Chen, Zhuo Chang, Shijia Gu, Yuanhang Zong, Zhang Mei, Shiyu Wang, Hezixiang Hezixiang, Hongzhi Chen, Jin Wei, Bin Cui
| Challenge: | et al., 2017) address domain-specific knowledge barriers, schemas complexity, and computational costs of large LLMs. |
| Approach: | They propose a domain-adapted Text2SQL system that addresses critical deployment challenges in professional fields. |
| Outcome: | The proposed system achieves 97% execution accuracy on real-world databases . it is faster than existing systems and has a higher performance than existing ones. |
MMDEND: Dendrite-Inspired Multi-Branch Multi-Compartment Parallel Spiking Neuron for Sequence Modeling (2025.acl-long)
Copied to clipboard
Kexin Wang, Yuhong Chou, Di Shang, Shijie Mei, Jiahong Zhang, Yanbin Huang, Man Yao, Bo Xu, Guoqi Li
| Challenge: | Vanilla spiking neurons are simplified from complex biological neurons with dendrites, soma, and synapses into single somatic compartments. |
| Approach: | They propose a multi-branch, multi-compartment parallel spiking dendritic neuron with a proportion-adjustable multi-branched structure that enables long-term temporal dependencies. |
| Outcome: | The proposed model achieves better long-sequence modeling capability with fewer parameters and lower energy consumption. |
Unifying Discrete and Continuous Representations for Unsupervised Paraphrase Generation (2023.emnlp-main)
Copied to clipboard
Mingfeng Xue, Dayiheng Liu, Wenqiang Lei, Jie Fu, Jian Lan, Mei Li, Baosong Yang, Jun Xie, Yidan Zhang, Dezhong Peng, Jiancheng Lv
| Challenge: | Existing unsupervised paraphrase generation methods require large-scale, manually annotated paraphrase datasets, which are labor-intensive to build. |
| Approach: | They propose a self-supervised pseudo-data construction method that generates diverse pseudo-paraphrases in distinct surface structures for a given sentence. |
| Outcome: | The proposed method generates diverse pseudo-paraphrases in distinct surface structures for a given sentence. |
Understanding Client Reactions in Online Mental Health Counseling (2023.acl-long)
Copied to clipboard
| Challenge: | Communication success relies heavily on reading participants’ reactions, but little research is on how listeners' reactions shape trajectories and outcomes of conversations. |
| Approach: | They propose to use client reactions to predict counseling outcomes by using an annotation framework that encompasses counselors’ strategies and client reaction behaviors. |
| Outcome: | The proposed framework can predict counselors' strategies and client reaction behaviors against a large-scale text-based counseling dataset. |
Logical Consistency is Vital: Neural-Symbolic Information Retrieval for Negative-Constraint Queries (2025.findings-acl)
Copied to clipboard
Ganlin Xu, Zhoujia Zhang, Wangyi Mei, Jiaqing Liang, Weijia Lu, Xiaodong Zhang, Zhifei Yang, Xiaofeng Ma, Yanghua Xiao, Deqing Yang
| Challenge: | Current dense retrieval methods compute similarities between dense vectors but overlook the real query intents. |
| Approach: | They propose a neuro-symbolic information retrieval method that leverages first-order logic to optimize the embeddings of naive natural language by considering the logical consistency between queries and documents. |
| Outcome: | The proposed method outperforms existing methods on negative-constraint queries under zero-shot and low-resource retrieval tasks. |
A Lightweight Mixture-of-Experts Neural Machine Translation Model with Stage-wise Training Strategy (2024.findings-naacl)
Copied to clipboard
| Challenge: | Using mixture-of-experts (MoE) to deal with language heterogeneity is a challenge in neural machine translation (NMT). |
| Approach: | They propose a lightweight MoE-based NMT model that is trained via an elaborate stage-wise training strategy. |
| Outcome: | The proposed model achieves stable improvements in translation tasks by introducing fewer extra parameters compared to baseline models. |
Dovetail: A CPU/GPU Heterogeneous Speculative Decoding for LLM inference (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) are demanding more memory and computational resources . however, these devices typically feature weaker GPUs and stronger CPUs . |
| Approach: | They propose a lossless inference acceleration method that leverages the characteristics of heterogeneous devices and the advantages of speculative decoding. |
| Outcome: | The proposed method achieves speedups ranging from 1.79 to 10.1 across different devices . it uses a draft model on the GPU to perform preliminary predictions, while a target model on CPU validates these outputs . |
iAgent: LLM Agent as a Shield between User and Recommender Systems (2025.findings-acl)
Copied to clipboard
Wujiang Xu, Yunxiao Shi, Zujie Liang, Xuying Ning, Kai Mei, Kun Wang, Xi Zhu, Min Xu, Yongfeng Zhang
| Challenge: | Traditional recommender systems focus on the user-platform paradigm, where users are directly exposed under the control of the platform's recommendation algorithms. |
| Approach: | They propose a user-agent-platform paradigm where agent serves as the protective shield between user and recommender system that enables indirect exposure. |
| Outcome: | The proposed model improves 16.6% over baselines on four datasets and mitigates echo chamber effects and reduces model bias in disadvantaged users. |
Exploring Concept Depth: How Large Language Models Acquire Knowledge and Concept at Different Layers? (2025.coling-main)
Copied to clipboard
Mingyu Jin, Qinkai Yu, Jingyuan Huang, Qingcheng Zeng, Zhenting Wang, Wenyue Hua, Haiyan Zhao, Kai Mei, Yanda Meng, Kaize Ding, Fan Yang, Mengnan Du, Yongfeng Zhang
| Challenge: | Large language models have shown remarkable performances across a wide range of tasks, but mechanisms by which they encode tasks of varying complexity remain poorly understood. |
| Approach: | They propose to explore the possibility that LLMs process concepts in different layers . they propose to categorize concepts based on their level of abstraction . |
| Outcome: | The proposed model can process complex concepts in shallow layers, the authors show . the proposed model could be used to prob complex tasks in shallow ones . |
MASSW: A New Dataset and Benchmark Tasks for AI-Assisted Scientific Workflows (2025.findings-naacl)
Copied to clipboard
Xingjian Zhang, Yutong Xie, Jin Huang, Jinge Ma, Zhaoying Pan, Qijia Liu, Ziyang Xiong, Tolga Ergen, Dongsub Shim, Honglak Lee, Qiaozhu Mei
| Challenge: | Scientific innovation is driven by detailed workflows, which include critical steps such as contextualizing literature, generating ideas, validating ideas, and planning new research. |
| Approach: | They propose to use large language models to extract five key aspects from scientific publications to optimize scientific workflows. |
| Outcome: | The proposed dataset includes more than 152,000 peer-reviewed publications from 17 leading computer science conferences spanning the past 50 years. |
PRewrite: Prompt Rewriting with Reinforcement Learning (2024.acl-short)
Copied to clipboard
| Challenge: | Prompt engineering is done manually in a trial-and-error ad-hoc fashion, authors say . |
| Approach: | They propose a method to rewrite an under-optimized prompt to a more effective prompt. |
| Outcome: | The proposed method rewrites an under-optimized prompt to a more effective prompt. |
Vulnerability of Text-to-Image Models to Prompt Template Stealing: A Differential Evolution Approach (2025.findings-acl)
Copied to clipboard
Yurong Wu, Fangwen Mu, Qiuhong Zhang, Jinjing Zhao, Xinrun Xu, Lingrui Mei, Yang Wu, Lin Shi, Junjie Wang, Zhiming Ding, Yiwei Wang
| Challenge: | Prompt trading has emerged as a significant intellectual property concern in recent years, where vendors entice users by showcasing sample images before selling prompt templates that can generate similar images. |
| Approach: | They propose a prompt-stealing benchmark consisting of 50 templates and 450 images organized into Easy and Hard difficulty levels. |
| Outcome: | The proposed method outperforms baseline methods with an average improvement of over 10%. |
Beyond Overlap Metrics: Rewarding Reasoning and Preferences for Faithful Multi-Role Dialogue Summarization (2026.findings-acl)
Copied to clipboard
| 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. |
HighMATH: Evaluating Math Reasoning of Large Language Models in Breadth and Depth (2025.findings-emnlp)
Copied to clipboard
Yan Liu, Minghui Zhang, Bojian Xiong, Yifan Xiao, Yinong Sun, Yating Mei, Longyu Zeng, Jingchao Yang, Yang Wang, Deyi Xiong
| Challenge: | a gap in math models' accuracy has been widened with the development of large language models (LLMs) . a new study aims to bridge this gap by evaluating a set of high-level math reasoning models . |
| Approach: | They propose to evaluate large language models on existing math benchmarks to bridge this gap . they collect 5,293 problems from Chinese senior high school mathematics exams . |
| Outcome: | The proposed model is based on o1-like models and a high-level model. |
CCIM: Cross-modal Cross-lingual Interactive Image Translation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing research on text image machine translation (TIMT) lacks recognized source language information resulting in a decrease in translation performance. |
| Approach: | They propose a cross-modal cross-lingual interactive model which incorporates source language information by synchronizing source and target language results. |
| Outcome: | The proposed model outperforms end-to-end models and has faster decoding speed with smaller model size than cascade models. |
Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approach (2024.emnlp-industry)
Copied to clipboard
| Challenge: | Recent LLMs like Gemini-1.5 and GPT-4 show exceptional capabilities to understand long contexts directly. |
| Approach: | They propose a method that routes queries to RAG or LC based on model self-reflection. |
| Outcome: | The proposed method significantly reduces the computation cost while maintaining a comparable performance to RAG. |
Defense against Prompt Injection Attacks via Mixture of Encodings (2025.naacl-short)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have emerged as a dominant approach for a wide range of NLP tasks, but external content embeds malicious instructions that manipulate the LLM’s output. |
| Approach: | They propose a mixture of encodings defense mechanism which utilizes multiple character encodes to degrade LLM performance on certain NLP tasks. |
| Outcome: | The proposed method achieves one of the lowest attack success rates under prompt injection attacks while maintaining high performance across all NLP tasks. |
AdaptFlow: Adaptive Workflow Optimization via Meta-Learning (2025.findings-emnlp)
Copied to clipboard
Runchuan Zhu, Bowen Jiang, Lingrui Mei, Fangkai Yang, Lu Wang, Haoxiang Gao, Fengshuo Bai, Pu Zhao, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang
| Challenge: | Existing approaches to large language models rely on static templates or manual workflows. |
| Approach: | AdaptFlow is a language-based meta-learning framework inspired by model-agnostic meta- learning. |
| Outcome: | AdaptFlow outperforms manual and automated workflows on question answering, code generation and mathematical reasoning benchmarks. |
Seeing Beyond Words: MatVQA for Challenging Visual-Scientific Reasoning in Materials Science (2026.findings-acl)
Copied to clipboard
| Challenge: | Multimodal Large Language Models (MLLMs) outperform existing benchmarks in both natural language and coding domains. |
| Approach: | They propose a scalable benchmark that integrates vision and language modalities to address this gap by eliminating textual shortcuts. |
| Outcome: | The new benchmark outperforms existing benchmarks in both natural language and coding domains. |
Long-range Sequence Modeling with Predictable Sparse Attention (2022.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to capture global context dependencies in sequence modeling suffer from quadratic complexity in time and memory usage. |
| Approach: | They propose an efficient Transformer architecture for fast long-range sequence modeling with a sparse attention matrix and a hidden state cross module. |
| Outcome: | The proposed architecture outperforms the standard multi-head attention and its variants in various long-sequence tasks with low computational costs. |
Bridging the Preference Gap between Retrievers and LLMs (2024.acl-long)
Copied to clipboard
| Challenge: | Existing studies on retrievers and LLMs treat them as separate components . a novel bridge model is proposed to optimize the relationship between the retriever and the LLM . |
| Approach: | They propose a framework that chains together supervised and reinforcement learning to train a bridge model that optimizes the connection between the retriever and the LLM. |
| Outcome: | Empirical results show that the proposed model optimizes the connection between the retriever and the LLM. |
ChildEval:WHEN LARGE LANGUAGE MODELS MEET CHILDREN’S PERSONALITIES (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have achieved remarkable success in effectively understanding and generating human language, leading to a revolutionary era in LLMs. |
| Approach: | They propose a benchmark to evaluate LLMs' ability to infer and follow child-centered preferences in long-context conversations. |
| Outcome: | The proposed benchmark spans five top-level and fourteen sub-level categories covering children’s daily lives and development. |