Papers by Yufei Li
Enabling Self-Improving Agents to Learn at Test Time With Human-In-The-Loop Guidance (2025.emnlp-industry)
Copied to clipboard
Yufei He, Ruoyu Li, Alex Chen, Yue Liu, Yulin Chen, Yuan Sui, Cheng Chen, Yi Zhu, Luca Luo, Frank Yang, Bryan Hooi
| Challenge: | Existing large language model (LLM) agents are unable to adapt to changing domain knowledge and rules. |
| Approach: | They propose an LLM agent framework that continuously learns updated domain knowledge at test time. |
| Outcome: | The proposed agent improves on a customer due diligence name screening task on . the agent learns updated domain knowledge at test time. |
Distantly-Supervised Joint Extraction with Noise-Robust Learning (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to identifying entity pairs and relations with a single model are noisy . Existing methods only consider one source of noise or make decisions using external knowledge . |
| Approach: | They propose a framework that aligns entity mentions with corresponding tags for joint extraction . they propose DENRL, which employs a lightweight transformer backbone for joint tagging . |
| Outcome: | The proposed framework outperforms baseline models on two benchmark datasets with better interpretability. |
Uncertainty-Aware Bootstrap Learning for Joint Extraction on Distantly-Supervised Data (2023.acl-short)
Copied to clipboard
| Challenge: | Existing approaches to extract entity pairs and their relations from labeled data are noisy and expensive. |
| Approach: | They propose a bootstrap learning approach that is motivated by intuition that the higher the uncertainty of an instance, the more likely the model confidence is inconsistent with the ground truths. |
| Outcome: | The proposed method outperforms baselines and related methods on two large datasets. |
Robustness via Referencing: Defending against Prompt Injection Attacks by Referencing the Executed Instruction (2026.findings-acl)
Copied to clipboard
Yulin Chen, Haoran Li, Yuan Sui, Yue Liu, Yufei He, Xiaoling Bai, Chi Fei, Li Yabo, Haozhe Ma, Yangqiu Song, Bryan Hooi
| Challenge: | Prompt injection attacks manipulate large language models (LLMs) by misleading them to deviate from the original input instructions and execute maliciously injected instructions. |
| Approach: | They propose a prompt injection defense method that suppresses the model's instruction-following tendencies rather than suppressing them. |
| Outcome: | The proposed method outperforms prompt-engineering-based approaches and fine-tuning methods and reduces the ASR to nearly 0% in some scenarios. |
CSMCIR: CoT-Enhanced Symmetric Alignment with Memory Bank for Composed Image Retrieval (2026.findings-acl)
Copied to clipboard
Zhipeng Qian, Zihan Liang, Yufei Ma, Ben Chen, Huangyu Dai, Yiwei Ma, Jiayi Ji, Chenyi Lei, Han Li, Xiaoshuai Sun
| Challenge: | Existing approaches to search for images using single-modality are limited by representation space fragmentation. |
| Approach: | They propose a unified representation framework that achieves efficient query-target alignment . they introduce a multi-level Chain-of-Thought prompting strategy that guides MLMs to generate discriminative, semantically compatible captions for target images . |
| Outcome: | The proposed framework achieves efficient query-target alignment through synergistic components. |
OpenEval: Benchmarking Chinese LLMs across Capability, Alignment and Safety (2024.acl-demos)
Copied to clipboard
Chuang Liu, Linhao Yu, Jiaxuan Li, Renren Jin, Yufei Huang, Ling Shi, Junhui Zhang, Xinmeng Ji, Tingting Cui, Liutao Liutao, Jinwang Song, Hongying Zan, Sun Li, Deyi Xiong
| Challenge: | a rapid development of Chinese large language models poses big challenges for efficient LLM evaluation. |
| Approach: | They propose an evaluation testbed that benchmarks Chinese LLMs across capability, alignment and safety. |
| Outcome: | The evaluation platform OpenEval benchmarks Chinese LLMs across capability, alignment and safety. |
Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack (2024.acl-long)
Copied to clipboard
| Challenge: | Recent developments in balancing usefulness and safety of large language models raise a critical question . current attacks, especially adversarial ones that manipulate malicious prompts, often aim to manipulate the input . |
| Approach: | They show that LLMs can effectively summarize malicious long documents but often refuse to translate them. |
| Outcome: | The findings highlight a vulnerability in LLMs that can't translate or summarize documents . the study focuses on LLM models, Gemini and GPT-4, which can' be exploited . |
Document Segmentation Matters for Retrieval-Augmented Generation (2025.findings-acl)
Copied to clipboard
Zhitong Wang, Cheng Gao, Chaojun Xiao, Yufei Huang, Shuzheng Si, Kangyang Luo, Yuzhuo Bai, Wenhao Li, Tangjian Duan, Chuancheng Lv, Guoshan Lu, Gang Chen, Fanchao Qi, Maosong Sun
| Challenge: | Existing rule-based chunking methods lead to suboptimal splits, where overly large chunks introduce irrelevant information and small chunks lack semantic coherence. |
| Approach: | They propose a method that leverages document summaries as pseudo-instructions to guide chunking by computing semantic similarity between sentences and the summary. |
| Outcome: | Experiments on multiple open-domain question-answering benchmarks show that PIC significantly improves retrieval accuracy (Hits@k) and end-to-end QA performance (Exact Match) without any additional training. |
Can Indirect Prompt Injection Attacks Be Detected and Removed? (2025.acl-long)
Copied to clipboard
| Challenge: | Recent studies have developed various detection mechanisms to protect against prompt injection attacks. |
| Approach: | They investigate the feasibility of detecting and removing indirect prompt injection attacks . they use two methods to evaluate their performance and train detection models . |
| Outcome: | The proposed method is based on a benchmark dataset and is available on github . it evaluates the performance of existing models and open-source detection models . |
REFFLY: Melody-Constrained Lyrics Editing Model (2025.naacl-long)
Copied to clipboard
| Challenge: | Automatic melody-to-lyric (M2L) generation aims to create lyrics that align with a given melody. |
| Approach: | They propose a framework for automatic melody-to-lyric generation that allows for a more flexible approach to creating lyrics from plain text. |
| Outcome: | The proposed framework outperforms baselines Lyra and GPT-4 in musicality and text quality. |
SHARE: a System for Hierarchical Assistive Recipe Editing (2022.emnlp-main)
Copied to clipboard
| Challenge: | Existing recipe websites do not provide options for users with dietary restrictions . a growing population follows some form of dietary restriction, with many people following it for a variety of reasons . |
| Approach: | They propose a system for hierarchical assistive recipe editing that performs simultaneous ingredient substitution before generating natural-language steps using the edited ingredients. |
| Outcome: | The proposed system can adapt a recipe to satisfy a user-specified dietary constraint. |
Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs’ Reasoning (2025.emnlp-main)
Copied to clipboard
Zezhong Wang, Xingshan Zeng, Weiwen Liu, Yufei Wang, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Qun Liu, Kam-Fai Wong
| Challenge: | Existing methods that use Chain-of-Thought suffer from path homogenization and inefficient use of intermediate results. |
| Approach: | They propose a framework that introduces checkpoints between reasoning steps to reduce path homogenization and create fault-tolerant mechanisms. |
| Outcome: | The proposed framework reduces path homogenization and creates fault-tolerant mechanism by utilizing high-quality intermediate results. |
White-Box Multi-Objective Adversarial Attack on Dialogue Generation (2023.acl-long)
Copied to clipboard
| Challenge: | Pre-trained transformers are popular in state-of-the-art dialogue generation systems . however, they are vulnerable to adversarial samples crafted by small and imperceptible perturbations. |
| Approach: | They propose a multi-objective attack method that balances two objectives: generation accuracy and length. |
| Outcome: | The proposed method significantly degrades state-of-the-art DG models with a higher success rate than traditional accuracy-based methods. |
Crowd Comparative Reasoning: Unlocking Comprehensive Evaluations for LLM-as-a-Judge (2025.acl-long)
Copied to clipboard
Qiyuan Zhang, Yufei Wang, Yuxin Jiang, Liangyou Li, Chuhan Wu, Yasheng Wang, Xin Jiang, Lifeng Shang, Ruiming Tang, Fuyuan Lyu, Chen Ma
| Challenge: | Existing methods rely on majority voting or criteria expansion to capture detailed and detailed details, often leading to incomplete outcomes. |
| Approach: | They propose a method which introduces additional crowd responses to compare with the candidate responses, thereby exposing deeper and more comprehensive details within the candidate answers. |
| Outcome: | Experiments show that the proposed method improves evaluation reliability and achieves an average gain of 6.7% across five benchmarks. |
GLTW: Joint Improved Graph Transformer and LLM via Three-Word Language for Knowledge Graph Completion (2025.findings-acl)
Copied to clipboard
Kangyang Luo, Yuzhuo Bai, Cheng Gao, Shuzheng Si, Zhu Liu, Yingli Shen, Zhitong Wang, Cunliang Kong, Wenhao Li, Yufei Huang, Ye Tian, Xuantang Xiong, Lei Han, Maosong Sun
| Challenge: | Existing knowledge graphs lack the ability to integrate structural information into LLMs and output predictions deterministically. |
| Approach: | They propose a method which encodes structural information of KGs and merges it with LLMs to enhance KGC performance. |
| Outcome: | The proposed method improves the performance of KG Completion datasets on KGs by integrating structural information with LLMs. |
Safety in Large Reasoning Models: A Survey (2025.findings-emnlp)
Copied to clipboard
Cheng Wang, Yue Liu, Baolong Bi, Duzhen Zhang, Zhong-Zhi Li, Yingwei Ma, Yufei He, Shengju Yu, Xinfeng Li, Junfeng Fang, Jiaheng Zhang, Bryan Hooi
| Challenge: | Large Reasoning Models (LRMs) have a high level of advanced reasoning capabilities, but they are vulnerable and vulnerable. |
| Approach: | This paper presents the first comprehensive survey of Large Reasoning Models . it explores the new safety risks, attacks, and defense strategies specific to LRMs based on reasoning . |
| Outcome: | The proposed study examines the safety and security risks of large reasoning models. |
Complementary Evidence Identification in Open-Domain Question Answering (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing approaches to QA that only measure the relevance between the question and each paragraph are not effective. |
| Approach: | They propose a method that learns vector representations of passages and models the sufficiency and diversity within the selected set, in addition to the relevance between the question and passages. |
| Outcome: | The proposed method significantly improves the accuracy of complementary evidence selection in open-domain question answering domain. |
Learning to Edit: Aligning LLMs with Knowledge Editing (2024.acl-long)
Copied to clipboard
Yuxin Jiang, Yufei Wang, Chuhan Wu, Wanjun Zhong, Xingshan Zeng, Jiahui Gao, Liangyou Li, Xin Jiang, Lifeng Shang, Ruiming Tang, Qun Liu, Wei Wang
| Challenge: | Existing knowledge editing techniques rely on memorizing updated knowledge, impeding LLMs from effectively combining the new knowledge with their inherent knowledge when answering questions. |
| Approach: | They propose a Learning to Edit framework that equips LLMs with the ability to apply updated knowledge to input questions through a two-phase process . |
| Outcome: | The proposed framework outperforms existing methods in knowledge editing tasks and compares it with four benchmarks and two LLM architectures. |
Knowledge Is Flat: A Seq2Seq Generative Framework for Various Knowledge Graph Completion (2022.coling-1)
Copied to clipboard
| Challenge: | Knowledge Graph Completion (KGC) has been extended to multiple knowledge graph (KG) structures, initiating new research directions, e.g. static KGC, temporal KGC and few-shot KGC. |
| Approach: | They propose a generative framework that could tackle different verbalizable graph structures by unifying the representation of KG facts into "flat" text. |
| Outcome: | The proposed framework outperforms many competitive baselines and sets new state-of-the-art performance on five benchmarks. |
Beyond Local vs. External: A Game-Theoretic Framework for Trustworthy Knowledge Acquisition (2026.findings-acl)
Copied to clipboard
| Challenge: | Cloud-hosted Large Language Models (LLMs) offer unmatched reasoning capabilities and dynamic knowledge, yet submitting raw queries to these services can expose sensitive user intent. |
| Approach: | They propose a framework that formulates the trade-off between knowledge utility and privacy as a strategic game. |
| Outcome: | The proposed framework reduces intent leakage while maintaining high-fidelity answer quality. |
Beyond Dialogue Time: Temporal Semantic Memory for Personalized LLM Agents (2026.findings-acl)
Copied to clipboard
Miao Su, Yucan Guo, Zhongni Hou, Long Bai, Zixuan Li, Yufei Zhang, Guojun Yin, Wei Lin, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng
| Challenge: | Existing methods focus on point-wise memory, losing durative information that captures persistent states and evolving patterns. |
| Approach: | They propose a memory framework that models semantic time for point-wise memory and supports the construction and utilization of durative memory. |
| Outcome: | Experiments on LongMemEval and LoCoMo show that the proposed method outperforms existing methods and achieves up to 12.2% improvement in accuracy. |
NILE: Internal Consistency Alignment in Large Language Models (2025.emnlp-main)
Copied to clipboard
Minda Hu, Qiyuan Zhang, Yufei Wang, Bowei He, Hongru Wang, Jingyan Zhou, Liangyou Li, Yasheng Wang, Chen Ma, Irwin King
| Challenge: | Recent advances show that the world knowledge in the Instruction Fine-Tuning (IFT) dataset, which is incompatible with LLMs’ internal knowledge, can greatly hurt the IFT performance. |
| Approach: | They propose a framework to optimize the effectiveness of IFT by carefully aligning the world and internal knowledge of LLMs. |
| Outcome: | The proposed framework can significantly improve performance across multiple LLM ability evaluation datasets. |
FollowBench: A Multi-level Fine-grained Constraints Following Benchmark for Large Language Models (2024.acl-long)
Copied to clipboard
Yuxin Jiang, Yufei Wang, Xingshan Zeng, Wanjun Zhong, Liangyou Li, Fei Mi, Lifeng Shang, Xin Jiang, Qun Liu, Wei Wang
| Challenge: | Existing benchmarks focus on evaluating pure response quality, rather than assessing whether the response follows constraints stated in the instruction. |
| Approach: | They propose a Multi-level Fine-grained Constraints Following Benchmark for Large Language Models that adds a single constraint to the initial instruction at each increased level. |
| Outcome: | The proposed model can follow instructions with more constraints, and is deemed to have better instruction-following ability. |
Paraphrase Generation as Unsupervised Machine Translation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing methods for paraphrase generation rely on labeled datasets or are limited in narrow domains. |
| Approach: | They propose a paradigm for paraphrase generation by treating the task as unsupervised machine translation based on pairs of unlabeled monolingual sentences. |
| Outcome: | The proposed paradigm can generate paraphrases on a large unlabeled monolingual corpus without relying on bilingual sentence pairs. |
Chain-of-Probe: Examining the Necessity and Accuracy of CoT Step-by-Step (2025.findings-naacl)
Copied to clipboard
Zezhong Wang, Xingshan Zeng, Weiwen Liu, Yufei Wang, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Qun Liu, Kam-Fai Wong
| Challenge: | Current research found the issue of Early Answering in large language models where the models already have an answer before generating the Chain-of-Thought (CoT). |
| Approach: | They propose a method to probe changes in confidence during the model’s reasoning and prioritize answers with correct reasoning among multiple candidates. |
| Outcome: | The proposed method reveals that in a significant number of question-answer cases, CoT appears to be unnecessary and this necessity correlates with the simplicity of the task, defined by the reasoning steps required. |
Com2 : A Causal-Guided Benchmark for Exploring Complex Commonsense Reasoning in Large Language Models (2025.acl-long)
Copied to clipboard
Kai Xiong, Xiao Ding, Yixin Cao, Yuxiong Yan, Li Du, Yufei Zhang, Jinglong Gao, Jiaqian Liu, Bing Qin, Ting Liu
| Challenge: | Existing works focus on complex tasks like math and code, while complex commonsense reasoning remains underexplored due to its uncertainty and lack of structure. |
| Approach: | They propose to build a benchmark for large language models based on complex commonsense reasoning based upon causal event graphs and causal theory. |
| Outcome: | The proposed benchmark combines a complex commonsense reasoning benchmark with a detective story to achieve a more challenging subset. |
BioGraphia: A LLM-Assisted Biological Pathway Graph Annotation Platform (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing methods for obtaining pathway information from biomedical literature rely on simplifying assumptions that limit their ability to capture true complexity of biological reactions. |
| Approach: | They propose a web-based platform to facilitate collaborative pathway graph annotation. |
| Outcome: | The platform supports multi-user collaboration with real-time monitoring, curation, and interactive pathway graph visualization. |
InfoGain-RAG: Boosting Retrieval-Augmented Generation through Document Information Gain-based Reranking and Filtering (2025.emnlp-main)
Copied to clipboard
Zihan Wang, Zihan Liang, Zhou Shao, Yufei Ma, Huangyu Dai, Ben Chen, Lingtao Mao, Chenyi Lei, Yuqing Ding, Han Li
| Challenge: | Retrieval-Augmented Generation (RAG) frameworks struggle with identifying whether retrieved documents meaningfully contribute to answer generation. |
| Approach: | They propose a document-related metric to quantify the contribution of retrieved documents to correct answer generation. |
| Outcome: | The proposed framework outperforms existing approaches on both single and multiple retrieval paradigms. |
Using Context-to-Vector with Graph Retrofitting to Improve Word Embeddings (2022.acl-long)
Copied to clipboard
| Challenge: | Contextualized embeddings are expensive and resource-demanding, hence environmentally unfriendly. |
| Approach: | They propose a method to convert contextualized embeddings from pre-trained models into static embeddables using synonym knowledge and weighted vector distribution. |
| Outcome: | The proposed method outperforms baseline embeddings by a large margin through extrinsic and intrinsic tasks. |
Let’s Negotiate! A Survey of Negotiation Dialogue Systems (2024.findings-eacl)
Copied to clipboard
Haolan Zhan, Yufei Wang, Zhuang Li, Tao Feng, Yuncheng Hua, Suraj Sharma, Lizhen Qu, Zhaleh Semnani Azad, Ingrid Zukerman, Reza Haf
| Challenge: | Recent research has focused on negotiation dialogue systems, but no systematic review of this task has been conducted. |
| Approach: | They propose to provide a systematic review of negotiation dialogue systems and to provide an overview of current research. |
| Outcome: | The proposed systems are based on the literature and are compared against existing systems. |
Dipping PLMs Sauce: Bridging Structure and Text for Effective Knowledge Graph Completion via Conditional Soft Prompting (2023.findings-acl)
Copied to clipboard
| Challenge: | Knowledge Graph Completion (KGC) often requires both KG structural and textual information to be effective. |
| Approach: | They propose a system which tunes the parameters of Conditional Soft Prompts generated by entities and relations representations to maintain a balance between textual and structural knowledge. |
| Outcome: | The proposed components outperform baseline models on three static and temporal benchmarks. |
MT-Eval: A Multi-Turn Capabilities Evaluation Benchmark for Large Language Models (2024.emnlp-main)
Copied to clipboard
Wai-Chung Kwan, Xingshan Zeng, Yuxin Jiang, Yufei Wang, Liangyou Li, Lifeng Shang, Xin Jiang, Qun Liu, Kam-Fai Wong
| Challenge: | Existing evaluation frameworks focus on single-turn evaluations, overlooking the models’ capabilities in multi-turn interactions. |
| Approach: | They propose a benchmark to evaluate the multi-turn conversational abilities of large language models (LLMs) by analyzing human-LLM conversations and constructing multi-turned queries for each category using GPT-4. |
| Outcome: | The proposed model outperforms open-source models in multi-turn tasks while retaining and recalling historical information. |