Papers by Zhuo Li
LaCo: Layer-wise Compensation for Pruned Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for predicting performance degradations of Large Language Models (LLMs) neglect the structural distortions caused by sparsity. |
| Approach: | They propose a framework that reorients the recovery paradigm from global adaptation to hierarchical representation alignment by sequentially optimizing each layer to reconstruct the model's hidden states. |
| Outcome: | The proposed framework surpasses parameter-efficient baselines in perplexity reduction and zero-shot reasoning. |
Advancing Zero-shot Text-to-Speech Intelligibility across Diverse Domains via Preference Alignment (2025.acl-long)
Copied to clipboard
| Challenge: | Existing zero-shot text-to-speech systems struggle in challenging scenarios such as tongue twisters, repeated words, code-switching, and cross-lingual synthesis. |
| Approach: | They propose a dataset that leverages preference alignment techniques to improve performance . they also extend the Direct Preference Optimization framework to accommodate diverse TTS architectures . |
| Outcome: | The proposed dataset improves intelligibility, similarity, and audio quality for multiple models across domains. |
Add-One-In: Incremental Sample Selection for Large Language Models via a Choice-Based Greedy Paradigm (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies focus on individual quality and do not assess the value of training data. |
| Approach: | They propose a choice-based sample selection framework that evaluates sample quality . they use LLMs to evaluate the value of each option during the selection process . |
| Outcome: | The proposed model outperforms the full dataset and recent studies on a larger medical dataset. |
Adaptive Hyper-parameter Learning for Deep Semantic Retrieval (2023.emnlp-industry)
Copied to clipboard
| Challenge: | Existing methods for deep semantic retrieval are highly sensitive to hyper-parameters . a novel adaptive metric learning method is proposed to overcome this limitation . |
| Approach: | They propose a method that adaptively obtains hyper-parameters without fixed or extra-trainable hyper-parmeters . they adopt a symmetric metric learning method to mitigate model collapse issues . |
| Outcome: | The proposed method outperforms existing methods on a real-world dataset and brings economic benefits. |
Atoxia: Red-teaming Large Language Models with Target Toxic Answers (2025.findings-naacl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are still vulnerable to generation safety vulnerabilities. |
| Approach: | They propose a method that A**tacks LLMs with target "toxi" given a particular harmful answer, the method generates a user query and a misleading answer opening to examine the internal defects of a given LLM. |
| Outcome: | The proposed method detects safety risks in open-source models and state-of-the-art models such as GPT-4o. |
Graph Enhanced Contrastive Learning for Radiology Findings Summarization (2022.acl-long)
Copied to clipboard
| Challenge: | Existing methods for automating impression generation have limited the relationship between extra knowledge and the original findings. |
| Approach: | They propose a framework for automating impression generation that exploits extra knowledge and original findings . they propose combining key words and their relations to extract critical information . |
| Outcome: | The proposed framework exploits extra knowledge and the original findings in an integrated way . the state-of-the-art results on two datasets confirm the effectiveness of the proposed method . |
HiSVD: Principled Low-Rank Approximation of LLMs via Hierarchical Modeling of Information Capacity and Spectral Structure (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for generating layer importance ignore the fine-grained influence of spectral distribution shape. |
| Approach: | They propose a hierarchical rank allocation framework with two stages to address this gap . they propose SVD-based lowrank approximation that exploits spectral heterogeneity . |
| Outcome: | Experiments show that HiSVD outperforms state-of-the-art methods on LLMs . |
Towards Reliable Large Audio Language Model (2025.findings-acl)
Copied to clipboard
Ziyang Ma, Xiquan Li, Yakun Song, Wenxi Chen, Chenpeng Du, Jian Wu, Yuanzhe Chen, Zhuo Chen, Yuping Wang, Yuxuan Wang, Xie Chen
| Challenge: | Recent advances in large audio language models (LALMs) have demonstrated impressive results and promising prospects in universal understanding and reasoning across speech, music, and general sound. |
| Approach: | They propose to use training-free and training-based methods to enhance LALM reliability to different extents. |
| Outcome: | The proposed methods improve the reliability of large audio language models to different extents. |
Decomposition for Enhancing Attention: Improving LLM-based Text-to-SQL through Workflow Paradigm (2024.findings-acl)
Copied to clipboard
Yuanzhen Xie, Xinzhou Jin, Tao Xie, Matrixmxlin Matrixmxlin, Liang Chen, Chenyun Yu, Cheng Lei, Chengxiang Zhuo, Bo Hu, Zang Li
| Challenge: | In-context learning of large-language models has achieved remarkable success in the field of natural language processing . however, the single-step chain-of-thought prompting approach faces challenges such as attention diffusion and inadequate performance in complex tasks like text-to-SQL. |
| Approach: | They propose a workflow paradigm method to enhance the attention and problem-solving scope of large-language models through decomposition. |
| Outcome: | The proposed method outperforms existing methods on three datasets and improves the upper limit of LLM-based approaches. |
Assistant-Guided Mitigation of Teacher Preference Bias in LLM-as-a-Judge (2025.findings-emnlp)
Copied to clipboard
| Challenge: | LLM-as-a-Judge uses large language models to evaluate the quality of LLM generated responses, but training proxy judge models using evaluation data generated by powerful teacher models introduces a critical yet previously overlooked issue: teacher preference bias. |
| Approach: | They propose a new setting that incorporates an additional assistant model, which is not biased toward the teacher model’s responses, to complement the training data. |
| Outcome: | The proposed model reduces teacher preference bias while maintaining strong performance across six evaluation benchmarks. |
Integrating Data Validation with Large Language Models for Regulation-Guided Tabular Anomaly Detection (2026.acl-long)
Copied to clipboard
| Challenge: | Existing tabular anomaly detection methods focus on detecting anomalies based on data distribution without considering regulatory compliance. |
| Approach: | They propose a task that leverages regulations to detect anomalies in tabular data . they also develop three new datasets to address this task . |
| Outcome: | The proposed method outperforms baselines on three new datasets. |
Self-Edit: Fault-Aware Code Editor for Code Generation (2023.acl-long)
Copied to clipboard
| Challenge: | Existing Large language models (LLMs) have low pass rates and accuracy on competitive programming tasks. |
| Approach: | They propose a generate-and-edit approach that uses execution results of generated code from LLMs to improve code quality on competitive programming tasks. |
| Outcome: | The proposed method improves pass@1 by 89% on APPS-dev, 31% on apps-test, and 48% on HumanEval over nine popular code generation LLMs with parameter sizes ranging from 110M to 175B. |
Med-SRAF: A Multi-Agent Framework for Medical Reasoning via Semantic Routing and Agentic Fusion (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing RAG methods suffer from a two-part problem: semantic drift and concatenation fallacy . et al.: rapid development of Large Language Models has led to a paradigm shift in artificial intelligence . |
| Approach: | They propose a multi-agent retrieval augmentation framework guided by medical domain knowledge to address these challenges. |
| Outcome: | The proposed framework outperforms existing general RAG baselines on five widely used medical benchmarks. |
Adversarial Speech Generation and Natural Speech Recovery for Speech Content Protection (2022.lrec-1)
Copied to clipboard
| Challenge: | Currently, researchers focus on how to protect the speaker's identifiable information, represented as voiceprint, contained in the speech. |
| Approach: | They propose a frame-by-frame adversarial speech generation system to protect speech . they build an adversarials-based method that converts adversarially generated speech to human speech. |
| Outcome: | The proposed method can encode and recover any sensitive audio, and it is easy to be conducted with publicly available speech recognition technology. |
On Robustness of Prompt-based Semantic Parsing with Large Pre-trained Language Model: An Empirical Study on Codex (2023.eacl-main)
Copied to clipboard
Terry Yue Zhuo, Zhuang Li, Yujin Huang, Fatemeh Shiri, Weiqing Wang, Gholamreza Haffari, Yuan-Fang Li
| Challenge: | Existing techniques for parsing natural-language utterances are vulnerable to adversarial attacks, requiring large amounts of labelled data and expensive human annotation. |
| Approach: | They propose to enhance the adversarial robustness of a prompt-based semantic parser based on a language model trained on code by constructing a set of demonstration examples. |
| Outcome: | The proposed method can be enhanced without significant amounts of labelled data or expensive human annotations on in-domain semantic parsing data. |
Not Just Plain Text! Fuel Document-Level Relation Extraction with Explicit Syntax Refinement and Subsentence Modeling (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Document-level relation extraction (DocRE) aims to identify semantic labels among entities within a document. |
| Approach: | They propose a document-level relation extraction framework that captures and exploits instructive information by adding extra syntactic information into text representations. |
| Outcome: | The proposed framework outperforms existing methods on three benchmark datasets. |
Benchmarking Vision-Language Models on Chinese Ancient Documents: From OCR to Knowledge Reasoning (2026.findings-acl)
Copied to clipboard
Haiyang Yu, Yuchuan Wu, Fan Shi, Jinghui Lu, Ke Niu, Xiaodong Ge, Minghan Zhuo, Jingqun Tang, Bin Li
| Challenge: | Existing document benchmarks focus on English printed texts or simplified Chinese . current vision-language models struggle with visual complexity and poor adaptability . |
| Approach: | They propose a benchmark to evaluate Chinese ancient documents' visual/linguistic complexity . ancient documents are valuable cultural heritage, but they face challenges in digitization and understanding . |
| Outcome: | the first benchmark for Chinese ancient documents evaluates VLMs from OCR to knowledge reasoning . ancient documents carry thousands of years of Chinese history and culture . traditional methods only scan images, while current models struggle with visual complexity . |
KBM: Delineating Knowledge Boundary for Adaptive Retrieval in Large Language Models (2025.findings-emnlp)
Copied to clipboard
Zhen Zhang, Xinyu Wang, Yong Jiang, Zile Qiao, Zhuo Chen, Guangyu Li, Feiteng Mu, Mengting Hu, Pengjun Xie, Fei Huang
| Challenge: | Retrieval-augmented generation (RAG) is employed to tackle these challenges . a Knowledge Boundary Model (KBM) is used to express the known/unknown of a given question . |
| Approach: | They propose a Knowledge Boundary Model to express the known/unknown of a given question . they find that not all questions need to trigger RAG to improve performance . |
| Outcome: | The proposed model reduces time and computational costs by retrieving parts of unknown knowledge . the proposed model can express the known/unknown of a given question and determine whether a RAG needs to be triggered . |
Prompting ChatGPT in MNER: Enhanced Multimodal Named Entity Recognition with Auxiliary Refined Knowledge (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to enhance textual entity prediction neglect the need for external knowledge or encounter high redundancy in the retrieved knowledge. |
| Approach: | They propose a framework that leverages ChatGPT as an implicit knowledge base and heuristically generates auxiliary knowledge for more efficient entity prediction. |
| Outcome: | The proposed framework outperforms state-of-the-art methods on two classic datasets and exhibits a stronger robustness and generalization capability. |
CoTEvol: Self-Evolving Chain-of-Thoughts for Data Synthesis in Mathematical Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to improve LLMs' reasoning abilities suffer from diminishing returns or high computing overhead. |
| Approach: | They propose a genetic evolutionary framework that casts CoT generation as a population-based search over reasoning trajectories. |
| Outcome: | The proposed framework improves correct-CoT synthesis success by over 30% and enhances structural diversity with markedly improved efficiency. |
From Implicit Exploration to Structured Reasoning: Guideline and Refinement for LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models rely on implicit exploration, which leads to unstable reasoning paths and lack of error correction. |
| Approach: | They propose a framework that shifts from implicit exploration to structured reasoning through guideline and refinement. |
| Outcome: | The proposed model outperforms strong baselines on the Big-Bench Hard benchmark. |
MARCH: Multi-Agent Reinforced Check for Hallucination (2026.acl-long)
Copied to clipboard
Zhuo Li, Yupeng Zhang, Pengyu Cheng, Jiajun Song, Mengyu Zhou, Hao Li, Shujie Hu, Yu Qin, null Erchao.zec, Xiaoxi Jiang, null Guanjunjiang
| Challenge: | Existing methods to detect hallucinations suffer from inherent confirmation bias, where the verifier inadvertently reproduces the errors of the original generation. |
| Approach: | They propose a framework that enforces rigorous factual alignment by leveraging deliberate *information asymmetry* by combining a pipeline of three specialized agents: a Solver, a Proposer, and a Checker. |
| Outcome: | Extensive experiments across hallucination benchmarks demonstrate that MARCH substantially reduces hallucinism rates. |
Self-Instructed Derived Prompt Generation Meets In-Context Learning: Unlocking New Potential of Black-Box LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Existing prompt refinement methods suffer from semantic inconsistencies and fail to maintain users’ real intent. |
| Approach: | They propose a self-instructed in-context learning framework that generates reliable derived prompts while keeping semantic consistency with original prompts. |
| Outcome: | The proposed framework generates better derived prompts and significantly enhances LLMs’ ability to deliver more effective responses. |
AD-LLM: Benchmarking Large Language Models for Anomaly Detection (2025.findings-acl)
Copied to clipboard
Tiankai Yang, Yi Nian, Li Li, Ruiyao Xu, Yuangang Li, Jiaqi Li, Zhuo Xiao, Xiyang Hu, Ryan A. Rossi, Kaize Ding, Xia Hu, Yue Zhao
| Challenge: | Anomaly detection (AD) is an important machine learning task with many real-world uses, including fraud detection, medical diagnosis, and industrial monitoring. |
| Approach: | They propose a benchmark that evaluates how large language models (LLMs) can help with NLP anomaly detection. |
| Outcome: | The proposed model can perform zero-shot detection without tasks-specific training, data augmentation and model selection, and it can suggest unsupervised AD models. |
Knowledgeable Preference Alignment for LLMs in Domain-specific Question Answering (2024.findings-acl)
Copied to clipboard
| Challenge: | Domain-specific question answering (QA) requires a comprehensive understanding of a specific domain to answer specialized questions. |
| Approach: | They propose a new alignment objective to align the LLM preference with different human preferences uniformly to optimize LLM performance in real-world, domain-specific QA settings. |
| Outcome: | The proposed pipeline is superior for real-scenario domain-specific question answering with LLMs. |
Intent-Driven Semantic ID Generation for Grounded Conversational News Recommendation (2026.acl-industry)
Copied to clipboard
| Challenge: | a new approach to news recommendation grounds each suggestion in a rapidly evolving article corpus while addressing implicit user intents that lack explicit retrievable keywords. |
| Approach: | They propose an intent-driven Semantic ID generation paradigm to address these challenges . they map diverse intents to hierarchical SID prefixes and then fuzzy-match them to current news pool . |
| Outcome: | The proposed model achieves 0% hallucination and 12.4% L1 match on a mainstream Chinese news platform. |
GigaSpeech 2: An Evolving, Large-Scale and Multi-domain ASR Corpus for Low-Resource Languages with Automated Crawling, Transcription and Refinement (2025.acl-long)
Copied to clipboard
Yifan Yang, Zheshu Song, Jianheng Zhuo, Mingyu Cui, Jinpeng Li, Bo Yang, Yexing Du, Ziyang Ma, Xunying Liu, Ziyuan Wang, Ke Li, Shuai Fan, Kai Yu, Wei-Qiang Zhang, Guoguo Chen, Xie Chen
| Challenge: | GigaSpeech 2 is a large-scale, multi-domain, multilingual speech recognition corpus for low-resource languages. |
| Approach: | They propose a large-scale, multi-domain, multilingual speech recognition corpus for low-resource languages and an automated pipeline for data crawling, transcription, and label refinement. |
| Outcome: | The proposed corpus reduces the word error rate for Thai, Indonesian, and Vietnamese on a realistic YouTube test set by 25% to 40% compared to Whisper large-v3. |
Skill Weaving: Efficient LLM Improvement via Modular Skillpacks (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) can specialize under fixed memory and inference budgets, but they struggle to achieve high performance across heterogeneous domains. |
| Approach: | They propose a modular improvement framework that partitions full capabilities of a general-purpose model into domain-specific delta modules that reorganize and refine the model's internal knowledge. |
| Outcome: | The proposed framework outperforms monolithic models on multi-task and agentic benchmarks and achieves up to 4 speedup. |
VoxMind: An End-to-End Agentic Spoken Dialogue System (2026.acl-long)
Copied to clipboard
Tianle Liang, Yifu Chen, Shengpeng Ji, Yijun Chen, Zhiyang Jia, Jingyu Lu, Fan Zhuo, Xueyi Pu, Yangzhuo Li, Zhou Zhao
| Challenge: | Existing research on end-to-end spoken dialogue models has focused on core perception and generation, with limited exploration of tool-augmented extensions. |
| Approach: | They propose a framework to equip end-to-end spoken dialogue models with comprehensive agentic abilities by leveraging a 470-hour AgentChat dataset. |
| Outcome: | The proposed framework outperforms Gemini-2.5-Pro on spoken agent tasks while maintaining general conversational quality. |
FACTUAL: A Benchmark for Faithful and Consistent Textual Scene Graph Parsing (2023.findings-acl)
Copied to clipboard
Zhuang Li, Yuyang Chai, Terry Yue Zhuo, Lizhen Qu, Gholamreza Haffari, Fei Li, Donghong Ji, Quan Hung Tran
| Challenge: | Existing parsers that convert image captions into scene graphs often suffer from errors and inconsistency. |
| Approach: | They propose a dataset that re-annotates image captions using a new intermediate representation called FACTUAL-MR and a metric to measure scene graph similarity. |
| Outcome: | The proposed parser outperforms existing parsers in terms of faithfulness and consistency on multiple benchmark datasets. |
Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Existing safety defenses for large language models fail to explicitly repel harmful patterns . Optimal transport (SOT) allows for safe fine-tuning without sacrificing safety . |
| Approach: | They propose a framework that reframes safe fine-tuning from instance-level filtering challenge to distribution-level alignment task grounded in Optimal Transport. |
| Outcome: | a new framework improves safety of large language models while maintaining competitive performance . the proposed framework reduces the risk of errors and improves model performance compared to baselines . |
NLP-ADBench: NLP Anomaly Detection Benchmark (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Anomaly detection (AD) is an important machine learning task, but its effectiveness in detecting harmful content, phishing attempts, and spam reviews is limited. |
| Approach: | They introduce NLP-ADBench, the most comprehensive NLP anomaly detection benchmark to date . it includes eight curated datasets and 19 state-of-the-art algorithms . |
| Outcome: | The NLP-ADBench benchmark includes 19 state-of-the-art methods and 8 curated datasets . no single model dominates across all datasets, indicating need for automated model selection . |
Self-Improvement Towards Pareto Optimality: Mitigating Preference Conflicts in Multi-Objective Alignment (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to optimize large language models with human preferences suffer from preference conflicts in the data. |
| Approach: | They propose to construct Pareto-optimal responses to resolve preference conflicts by using a self-improving DPO framework that enables LLMs to self-generate and select Paret-optimized responses. |
| Outcome: | The proposed framework achieves superior Pareto Front performance over baselines on two datasets. |
APLOT: Robust Reward Modeling via Adaptive Preference Learning with Optimal Transport (2025.emnlp-main)
Copied to clipboard
| Challenge: | Experimental results show that RLHF improves performance of Large Language Models . BT-based RMs struggle to distinguish between similar preference responses . |
| Approach: | They propose to enhance BT-based reward models by using an adaptive margin mechanism . they use semantic similarity and reward-predicted reward differences to adjust focus . |
| Outcome: | Experimental results show that the proposed method outperforms existing methods in both in-distribution and OOD settings. |
Self-Improvement Programming for Temporal Knowledge Graph Question Answering (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods implicitly model time constraints by learning time-aware embeddings of questions and candidate answers, which is far from understanding the question comprehensively. |
| Approach: | They propose a temporal-based temporal programming method that leverages the in-context learning ability of Large Language Models to understand combinatory time constraints in questions. |
| Outcome: | The proposed method outperforms existing methods on multiTQ and CronQuestions datasets and is highly efficient on multi-level questions. |
Breaking the Hourglass Phenomenon of Residual Quantization: Enhancing the Upper Bound of Generative Retrieval (2024.emnlp-industry)
Copied to clipboard
Zhirui Kuai, Zuxu Chen, Huimu Wang, Mingming Li, Dadong Miao, Wang Binbin, Xusong Chen, Li Kuang, Yuxing Han, Jiaxing Wang, Guoyu Tang, Lin Liu, Songlin Wang, Jingwei Zhuo
| Challenge: | Generative retrieval (GR) is a transformative paradigm in search and recommender systems . however, data sparsity and long-tailed distribution hinder the full utilization of GR . |
| Approach: | They propose a method to reduce the "Hourglass" phenomenon in RQ-SID where codebook tokens become overly concentrated. |
| Outcome: | The proposed methods improve retrieval efficiency and generalization capabilities. |
PEMT: Multi-Task Correlation Guided Mixture-of-Experts Enables Parameter-Efficient Transfer Learning (2024.findings-acl)
Copied to clipboard
| Challenge: | Parameter-efficient fine-tuning (PEFT) is an effective method for adapting pre-trained language models to various tasks efficiently. |
| Approach: | They propose a parameter-efficient fine-tuning framework that captures transferable knowledge as a weighted combination of adapters trained on source tasks. |
| Outcome: | The proposed method yields stable improvements over full fine-tuning and knowledge transferring methods on a broad range of tasks over 17 datasets. |
WavAlign: Enhancing Intelligence and Expressiveness in Spoken Dialogue Models via Adaptive Hybrid Post-Training (2026.findings-acl)
Copied to clipboard
Yifu Chen, Shengpeng Ji, Qian Chen, Tianle Liang, Yangzhuo Li, Ziqing Wang, Wen Wang, Jingyu Lu, Haoxiao Wang, Xueyi Pu, Fan Zhuo, Zhou Zhao
| Challenge: | End-to-end spoken dialogue models have higher potential ceiling in expressiveness and perceptual ability than cascaded systems. |
| Approach: | They propose a modality-aware adaptive post-training recipe that constrains preference updates to the semantic channel and improves acoustic behavior via explicit anchoring. |
| Outcome: | The proposed model improves speech quality and expressiveness across spoken dialogue benchmarks and architectures. |
Psyche-R1: Towards Reliable Psychological LLMs through Unified Empathy, Expertise, and Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Recent reasoning-augmented LLMs have demonstrated impressive capabilities across a wide range of domains owing to their exceptional text understanding capabilities. |
| Approach: | They propose a Chinese psychological LLM that integrates empathy, psychological expertise, and reasoning. |
| Outcome: | The proposed model produces over 75k high-quality psychological questions paired with detailed rationales, generated through and iterative prompt-rationale optimization procedure, along with 73k empathetic dialogues. |
TROJail: Trajectory-Level Optimization for Multi-Turn Large Language Model Jailbreaks with Process Rewards (2026.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to training multi-turn attackers to probe model safety vulnerabilities rely on turn-level optimization, which is insufficient for learning long-term attack strategies. |
| Approach: | They propose a multi-turn reinforcement learning problem that optimizes the harmfulness of the final-turn response as the outcome reward. |
| Outcome: | The proposed approach improves attack success rates across multiple models and benchmarks, highlighting the effectiveness of the proposed approach. |