Papers by Hui Wu
An Investigation of LLMs’ Inefficacy in Understanding Converse Relations (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing benchmarks for Large Language Models (LLMs) follow the data distribution of pre-training data. |
| Approach: | They propose a benchmark ConvRe focusing on converse relations which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. |
| Outcome: | The proposed benchmark focuses on converse relations, which contains 17 relations and 1240 triples extracted from popular knowledge graph completion datasets. |
From Adversarial Arms Race to Model-centric Evaluation: Motivating a Unified Automatic Robustness Evaluation Framework (2023.findings-acl)
Copied to clipboard
Yangyi Chen, Hongcheng Gao, Ganqu Cui, Lifan Yuan, Dehan Kong, Hanlu Wu, Ning Shi, Bo Yuan, Longtao Huang, Hui Xue, Zhiyuan Liu, Maosong Sun, Heng Ji
| Challenge: | Existing models of robustness evaluation are incomprehensive, impractical, and invalid . |
| Approach: | They propose a framework for automatic robustness evaluation that shifts towards model-centric evaluation to further exploit the advantages of adversarial attacks. |
| Outcome: | The proposed framework is based on a model-centric evaluation protocol and a robustness evaluation protocol. |
DiVE: Decoupling Intra-layer Visual Evidence for Mitigating Hallucinations in Large Vision-Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing decoding-based approaches do not explicitly decouple visual evidence from mixed vision–language representations. |
| Approach: | They propose to decouple visual evidence from mixed vision–language representations by dynamically identifying layers enriched with visual information and performing intra-layer decoupling to extract aggregated visual evidence. |
| Outcome: | Experiments show that DiVE achieves state-of-the-art performance on multiple benchmarks. |
Legal Fact Prediction: The Missing Piece in Legal Judgment Prediction (2025.emnlp-main)
Copied to clipboard
Junkai Liu, Yujie Tong, Hui Huang, Bowen Zheng, Yiran Hu, Peicheng Wu, Chuan Xiao, Makoto Onizuka, Muyun Yang, Shuyuan Zheng
| Challenge: | Existing studies use legal facts to predict judgments, but legal facts are difficult to obtain in early stages of litigation. |
| Approach: | They propose a legal fact prediction task that takes evidence from trial as input to make predictions in the absence of ground-truth legal facts. |
| Outcome: | The proposed task can predict court rulings without ground-truth legal facts . the first benchmark dataset, LFPBench, is used to evaluate the task . |
A Holistic Approach to Reference-Free Evaluation of Machine Translation (2023.acl-short)
Copied to clipboard
| Challenge: | Traditional machine translation evaluation relies on reference written by humans . reference-free evaluation gets rid of labor-intensive annotations, which can pivot easily to new domains . |
| Approach: | They propose a reference-free evaluation approach that characterizes evaluation as two aspects: fluency and faithfulness. |
| Outcome: | The proposed approach outperforms SOTA reference-fee metrics on machine translation datasets. |
Consistent Prototype Learning for Few-Shot Continual Relation Extraction (2023.acl-long)
Copied to clipboard
| Challenge: | Existing methods for few-shot continual relation extraction are overfitting memory samples, resulting in insufficient activation of old relations and limited ability to handle confusion of similar classes. |
| Approach: | They propose a few-shot continual relation extraction task that uses memory-enhanced modules to train a model on incrementally few-shot data to avoid forgetting old relations. |
| Outcome: | The proposed method outperforms existing methods on two commonly-used datasets. |
Divide-Verify-Refine: Can LLMs Self-align with Complex Instructions? (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing research shows LLMs struggle with complex instructions involving multiple constraints. |
| Approach: | They propose a framework to divide complex instructions into single constraints and prepare appropriate tools to verify responses. |
| Outcome: | The proposed framework doubles Llama3.1-8B’s constraint adherence and triples Mistral-7B’ s performance. |
Beyond Yes and No: Improving Zero-Shot LLM Rankers via Scoring Fine-Grained Relevance Labels (2024.naacl-short)
Copied to clipboard
| Challenge: | Existing pointwise LLMs provide noisy or biased answers for documents that are partially relevant to the query. |
| Approach: | They propose to incorporate fine-grained relevance labels into the LLM prompt . they propose to better differentiate between documents with different levels of relevance . |
| Outcome: | The proposed model can differentiate between documents with different levels of relevance to the query and derive a more accurate ranking. |
Synchronous Dual Network with Cross-Type Attention for Joint Entity and Relation Extraction (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on joint entity and relation extraction fail to fully utilize the interdependence between entity types and relation types. |
| Approach: | They propose a synchronous dual network with cross-type attention via separately and interactively considering the entity types and relation types. |
| Outcome: | The proposed model achieves state-of-the-art on NYT and WebNLG datasets. |
Anti-Length Shift: Dynamic Outlier Truncation for Training Efficient Reasoning Models (2026.acl-long)
Copied to clipboard
Wei Wu, Liyi Chen, Congxi Xiao, Tianfu Wang, Qimeng Wang, Chengqiang Lu, Yan Gao, null Yiwu, Yao Hu, Hui Xiong
| Challenge: | Existing efficient reasoning methods rely on explicit length penalties for excessive verbosity on simple queries. |
| Approach: | They propose a training-time intervention that selectively suppresses redundant tokens . they find length shift occurs when models generate unnecessary reasoning on trivial inputs - a phenomenon that is often unexplored . |
| Outcome: | The proposed method reduces inference token usage by 78% while increasing accuracy compared to the initial policy and surpasses state-of-the-art efficient reasoning methods. |
Separation and Fusion: A Novel Multiple Token Linking Model for Event Argument Extraction (2024.naacl-long)
Copied to clipboard
Jing Xu, Dandan Song, Siu Hui, Zhijing Wu, Meihuizi Jia, Hao Wang, Yanru Zhou, Changzhi Zhou, Ziyi Yang
| Challenge: | Existing methods for event argument extraction (EAE) lack cross-event information and require longer role sequences . et al. (2017): outperforms state-of-the-art methods for EE. |
| Approach: | They propose a separation-and-fusion paradigm to separate the acquisition of cross-event information and fuse it into the argument extraction of a target event. |
| Outcome: | The proposed model outperforms the state-of-the-art models on four widely used datasets. |
Many Heads Are Better Than One: Improved Scientific Idea Generation by A LLM-Based Multi-Agent System (2025.acl-long)
Copied to clipboard
Haoyang Su, Renqi Chen, Shixiang Tang, Zhenfei Yin, Xinzhe Zheng, Jinzhe Li, Biqing Qi, Qi Wu, Hui Li, Wanli Ouyang, Philip Torr, Bowen Zhou, Nanqing Dong
| Challenge: | Recent AI methods have shown promise in tasks such as hypothesis generation and experimental design, but they fail to replicate the collaborative nature of real-world scientific practices. |
| Approach: | They propose a virtual scientific system that mimics the collaborative nature of scientific research by organizing a team of agents to generate, evaluate, and refine research ideas. |
| Outcome: | The proposed system outperforms the state-of-the-art method in producing new scientific ideas and offers valuable insights to guide future research. |
STELLA: A Multimodal LLM for Protein Functional Annotation via Unified Sequence-Structure Encoding (2026.findings-acl)
Copied to clipboard
Hongwang Xiao, Wenjun Lin, Xi Chen, Hui Wang, Kai Chen, Jiashan Li, Yuancheng Sun, Sicheng Dai, Boya Wu, Qiwei Ye
| Challenge: | a multimodal protein language model (LLM) integrates sequence, structure, and function into functional annotation. |
| Approach: | They propose a multimodal protein language model that synergistically aligns bimodal representations with the textual modality to advance protein functional annotation. |
| Outcome: | The proposed model synergizes bimodal representations with the textual modality to advance protein functional annotation. |
FreebaseQA: A New Factoid QA Data Set Matching Trivia-Style Question-Answer Pairs with Freebase (N19-1)
Copied to clipboard
| Challenge: | Using FreebaseQA, we can generate over 54K matches from about 28K unique questions with minimal cost. |
| Approach: | They propose a data set for open-domain factoid question answering tasks over structured knowledge bases, like Freebase, using a combination of trivia-type question-answer pairs and subject-predicate-object triples. |
| Outcome: | The proposed data set generates 54K matches from 28K unique questions with minimal cost. |
Iterative Nearest Neighbour Machine Translation for Unsupervised Domain Adaptation (2023.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for supervised domain adaptation of machine translation focus on fine-tuning, which is non-extensible. |
| Approach: | They propose to perform unsupervised domain adaptation in a non-parametric manner by using in-domain monolingual data and performing nearest neighbour inference on both forward and backward directions. |
| Outcome: | The proposed method significantly improves the in-domain translation performance and achieves state-of-the-art results among non-parametric methods. |
Select2Reason: Efficient Instruction-Tuning Data Selection for Long-CoT Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large reasoning models exhibit human-like behaviors such as exploration, verification, reflection, and correction. |
| Approach: | They propose a supervised fine-tuning framework for long chain-of-thoughts reasoning . they leverage a difficulty-aware reward model to estimate the learning value of questions . |
| Outcome: | The proposed framework performs fine-tuning on large reasoning models on 10% of the data selected. |
ClaimGen-CN: A Large-scale Chinese Dataset for Legal Claim Generation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Currently, legal claims are not being used by non-professionals. |
| Approach: | They construct a dataset for Chinese legal claim generation task and then use it to evaluate the generated claims. |
| Outcome: | The proposed dataset is the first for the Chinese legal claim generation task and will be made publicly available. |
TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection (2025.emnlp-main)
Copied to clipboard
| Challenge: | Rapid advances in Large Language Models have spurred demand for processing extended context sequences . however, performance degradation due to sequence lengths out-of-distribution and excessively long inference times are limiting LLMs in long-context scenarios. |
| Approach: | They propose a training-free method for efficient and accurate long-context inference . they selectively involves a few critical KV cache tokens in attention calculation . |
| Outcome: | The proposed method speeds up attention computation and accelerates inference time while reducing selection overhead. |
SEC-FinTables: Evaluating Large Language Models for Detecting Logical Inconsistencies on Tabular Data (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models are increasingly deployed in high-stakes domains where logical inconsistencies are unrecognized. |
| Approach: | They propose a benchmarking system that decomposes inconsistency detection into granular subtasks and a protocol that decompiles it into subtask. |
| Outcome: | The proposed model decomposes inconsistencies into subtasks and identifies them in 103,395 real-world and error-injected table instances. |
Improving Translation Quality Estimation with Bias Mitigation (2023.acl-long)
Copied to clipboard
| Challenge: | State-of-the-art translation Quality Estimation models are biased, relying on monolingual features while ignoring the bilingual semantic alignment. |
| Approach: | They propose a method to mitigate the bias of translation quality estimation models by contrastive learning between clean and noisy sentence pairs. |
| Outcome: | The proposed method improves the estimation performance while mitigating the bias. |
Prompting Large Language Models to Tackle the Full Software Development Lifecycle: A Case Study (2025.coling-main)
Copied to clipboard
Bowen Li, Wenhan Wu, Ziwei Tang, Lin Shi, John Yang, Jinyang Li, Shunyu Yao, Chen Qian, Binyuan Hui, Qicheng Zhang, Zhiyin Yu, He Du, Ping Yang, Dahua Lin, Chao Peng, Kai Chen
| Challenge: | Existing benchmarks focused on simplified or isolated aspects of coding, ignoring the full spectrum of programming challenges. |
| Approach: | They propose a case study that examines the performance of large language models across the entire software development lifecycle with four programming languages, multiple domains, and carefully designed and verified metrics for each task. |
| Outcome: | The proposed model performs across the entire software development lifecycle, including design, environment setup, implementation, acceptance testing, and unit testing. |
HFT: Half Fine-Tuning for Large Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) with one or more fine-tuning phases can unlock various capabilities, but can be catastrophic forgetting during sequential training. |
| Approach: | They propose a method to regularly reset partial parameters to mitigate forgetting issues by using half fine-tuning instead of full fine-uning. |
| Outcome: | The proposed approach reduces the risk of catastrophic forgetting during training and the parametric knowledge lost during training may be overwhelmed by incoming training data. |
Benchmarking Hallucination in Large Language Models Based on Unanswerable Math Word Problem (2024.lrec-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) are highly effective in various natural language processing tasks, but can produce unreliable conjectures in ambiguous contexts, which is known as hallucination. |
| Approach: | They propose a method to evaluate LLM hallucination in Question Answering based on the unanswerable math word problem (UMWP) . they combine text similarity and mathematical expression detection to determine whether LLM considers the question unanswered. |
| Outcome: | The proposed method combines text similarity and mathematical expression detection to determine whether the LLM considers the question unanswerable. |
Adversarial Soft Prompt Tuning for Cross-Domain Sentiment Analysis (2022.acl-long)
Copied to clipboard
| Challenge: | Existing prompt tuning methods for cross-domain sentiment analysis have been underutilized due to domain discrepancy in the token distributions. |
| Approach: | They propose a new method to model cross-domain sentiment analysis using pre-trained language models by using soft prompts instead of hard templates. |
| Outcome: | The proposed method achieves state-of-the-art results on a publicly available sentiment analysis dataset. |
Qwen2.5-xCoder: Multi-Agent Collaboration for Multilingual Code Instruction Tuning (2025.acl-long)
Copied to clipboard
Jian Yang, Wei Zhang, Yibo Miao, Shanghaoran Quan, Zhenhe Wu, Qiyao Peng, Liqun Yang, Tianyu Liu, Zeyu Cui, Binyuan Hui, Junyang Lin
| Challenge: | Existing methods to train code LLMs view each programming language in isolation . experimental results show that Qwen2.5-xCoder can bridge the gap between different programming languages . |
| Approach: | They propose a framework that allows agents to collaborate to enhance multilingual instruction tuning for code LLMs. |
| Outcome: | Experimental results show that Qwen2.5-xCoder can transfer knowledge efficiently and effectively between languages. |
Large Language Models are Effective Text Rankers with Pairwise Ranking Prompting (2024.findings-naacl)
Copied to clipboard
Zhen Qin, Rolf Jagerman, Kai Hui, Honglei Zhuang, Junru Wu, Le Yan, Jiaming Shen, Tianqi Liu, Jialu Liu, Donald Metzler, Xuanhui Wang, Michael Bendersky
| Challenge: | Existing methods to rank documents using large language models do not understand these challenging ranking formulations. |
| Approach: | They propose to use Pairwise Ranking Prompting to improve ranking performance . they propose to outperform fine-tuned baseline rankers on benchmark datasets . |
| Outcome: | The proposed technique outperforms supervised baselines on benchmark datasets and outperformed other LLM-based solutions by over 10% on average. |
DREAM: Disentangling Risks to Enhance Safety Alignment in Multimodal Large Language Models (2025.naacl-long)
Copied to clipboard
Jianyu Liu, Hangyu Guo, Ranjie Duan, Xingyuan Bu, Yancheng He, Shilong Li, Hui Huang, Jiaheng Liu, Yucheng Wang, Chenchen Jing, Xingwei Qu, Xiao Zhang, Pei Wang, Yanan Wu, Jihao Gu, Yangguang Li, Jianke Zhu
| Challenge: | Multimodal Large Language Models (MLLMs) pose unique safety challenges due to their integration of visual and textual data. |
| Approach: | They propose a method to disentangle risks through step-by-step reasoning within multimodal inputs. |
| Outcome: | The proposed approach improves safety alignment in MLLMs by fine-tuning and iterative Reinforcement Learning from AI feedback. |
Exploring Dynamic Selection of Branch Expansion Orders for Code Generation (2021.acl-long)
Copied to clipboard
| Challenge: | Existing code generation models model abstract syntax tree (AST) but not suitable for all multi-branch nodes. |
| Approach: | They propose to equip a Seq2Tree model with a branch selector to determine optimal expansion orders for multi-branch nodes. |
| Outcome: | The proposed model can determine optimal expansion orders of branches for multi-branch nodes. |
LLM Sensitivity Evaluation Framework for Clinical Diagnosis (2025.coling-main)
Copied to clipboard
| Challenge: | Existing studies on the sensitivity of Large Language Models (LLMs) to irrelevant contexts neglect the importance of key information. |
| Approach: | They investigate the sensitivity of large language models to key medical information by introducing different perturbation strategies to investigate their sensitivity. |
| Outcome: | The proposed models are based on three LLMs, namely GPT-3.5, GPT-4, Gemini, Claude3 and LLaMA2-7b, and demonstrate their reliability and sensitivity to medical information. |
Monte Carlo Tree Search Based Prompt Autogeneration for Jailbreak Attacks against LLMs (2025.coling-main)
Copied to clipboard
| Challenge: | Jailbreak attacks craft specific prompts or append adversarial suffixes to prompts, thereby inducing language models to generate harmful or unethical content and bypassing the model’s safety guardrails. |
| Approach: | They propose a Monte Carlo Tree Search (MCTS) based Prompt Auto-generation (MPA) method to generate adversarial suffixes for valid jailbreak attacks. |
| Outcome: | The proposed method outperforms existing methods on open-source and closed-source models and shows that it can generate harmful responses. |
DebugBench: Evaluating Debugging Capability of Large Language Models (2024.findings-acl)
Copied to clipboard
Runchu Tian, Yining Ye, Yujia Qin, Xin Cong, Yankai Lin, Yinxu Pan, Yesai Wu, Hui Haotian, Liu Weichuan, Zhiyuan Liu, Maosong Sun
| Challenge: | Large language models (LLMs) have demonstrated exceptional coding capabilities, but their debugging capabilities remain relatively unexplored. |
| Approach: | They propose a debugging benchmark consisting of 4,253 LLMs with four major bug categories and 18 minor types in C++, Java, and Python. |
| Outcome: | The proposed benchmark covers four major bug categories and 18 minor types in C++, Java, and Python. |
MoEC: A Memory-Routed Mixture-of-Experts Controller for Adaptive Minecraft Control (2026.acl-long)
Copied to clipboard
| Challenge: | Existing systems rely on a monolithic policy to execute subgoals across varying contexts, causing inconsistent outcomes and scaling only partially mitigates. |
| Approach: | They propose a memory-routed mixtureof-experts controller for Adaptive Minecraft Control that routes via a subgoal-indexed expert memory and regulates capacity through failure-triggered expert growth and redundancy-aware consolidation. |
| Outcome: | The proposed controller shows significant gains in adaptability, robustness, and execution consistency over strong baselines. |
Upcycling Instruction Tuning from Dense to Mixture-of-Experts via Parameter Merging (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for tuning large language models from dense to MoE face significant data requirements and require large-scale post-training. |
| Approach: | They propose an upcycling instruction tuning approach for tuning a dense pre-trained model into a MoE instruction model using genetic algorithm and parameter merging. |
| Outcome: | The proposed approach improves the performance of large language models with a small amount of seed data and improves their scaling. |
A.S.E: A Repository-Level Benchmark for Evaluating Security in AI-Generated Code (2026.findings-acl)
Copied to clipboard
Keke Lian, Wang Bin, Lei Zhang, Libo Chen, Junjie Wang, Ziming Zhao, Yujiu Yang, Miaoqian Lin, Haotong Duan, Haoran Zhao, Shuang Liao, Mingda Guo, Quan Jiazheng, Yilu Zhong, Chenhao He, Chen Zichuan, Jie Wu, Haoling Li, Zhaoxuan Li, Jiongchi Yu, Hui LI, Dong Zhang
| Challenge: | Existing security evaluation benchmarks lack relevance to real-world AI programming tasks . current LLMs struggle with secure coding, research shows . |
| Approach: | They propose a repository-level evaluation benchmark to assess security of AI-generated code. |
| Outcome: | The proposed framework mirrors real-world AI programming tasks and offers valuable insights into the state of AI code generation. |
Conic10K: A Challenging Math Problem Understanding and Reasoning Dataset (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing benchmarks or datasets require only a few steps of reasoning, making it difficult to analyse AI’s behaviour with reference to different problems within a specific topic in detail. |
| Approach: | They propose a conic10K math problem dataset that requires only a few steps of reasoning to be analysed. |
| Outcome: | The proposed dataset shows that existing language models exhibit weak performance on complex reasoning. |