Papers by Xiang Cheng

ChemVLR: Prioritizing Reasoning in Perception for Chemical Vision-Language Understanding (2026.findings-acl)

Challenge:	Existing work on temporal sentence grounding rely on expensive video-query paired annotations . despite this, there are no ground-truth annotations in the current work .
Approach:	They propose to use paired video-query and segment boundary annotations to generate temporal sentence grounding without training.
Outcome:	The proposed model outperforms existing unsupervised methods and beats supervised ones on two challenging datasets.

Copied to clipboard

Xuanle Zhao, Xinyuan Cai, Xiang Cheng, Xiuyi Chen, Bo XU

Simple but Challenging: Natural Language Inference Models Fail on Simple Sentences (2022.findings-emnlp)

Challenge:	Currently, vision-Language Models are optimized for direct visual question-answering tasks.
Approach:	They propose a visual-language-based VLM that prioritizes reasoning within the perception process.
Outcome:	The proposed model outperforms existing models and domain-specific open-source models in the chemical domain.

Copied to clipboard

Cheng Luo, Wei Liu, Jieyu Lin, Jiajie Zou, Ming Xiang, Nai Ding

Atoxia: Red-teaming Large Language Models with Target Toxic Answers (2025.findings-naacl)

Challenge:	Natural language inference (NLI) tasks are difficult to perform on large datasets . a small number of simple sentences can improve model performance, authors say .
Approach:	They propose to use syntactically simple sentences to test the inference ability of NLI models.
Outcome:	The proposed set of simple sentences shows that the models fine-tuned on MNLI and SNLI perform poorly on Simple Pair.

Copied to clipboard

Yuhao Du, Zhuo Li, Pengyu Cheng, Xiang Wan, Anningzhe Gao

Call Me When Necessary: LLMs can Efficiently and Faithfully Reason over Structured Environments (2024.findings-acl)

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.

Copied to clipboard

Sitao Cheng, Ziyuan Zhuang, Yong Xu, Fangkai Yang, Chaoyun Zhang, Xiaoting Qin, Xiang Huang, Ling Chen, Qingwei Lin, Dongmei Zhang, Saravan Rajmohan, Qi Zhang

Reasoning

Challenge:	Large Language Models (LLMs) have shown potential in reasoning over structured environments, e.g., knowledge graphs and tables.
Approach:	They propose a framework that allows LLMs to efficiently and faithfully reason over structured environments.
Outcome:	The proposed framework surpasses state-of-the-art fine-tuned methods on three KGQA and two TableQA datasets and surpasse CWQ and WTQ methods.

From Implicit Graph Encoding to Explicit Evidence: A Training-Free LLM Framework for Temporal Knowledge Graph Reasoning (2026.findings-acl)

Copied to clipboard

Guo Tang, Ke Cheng, Huiming Fan, Heng Chang, Wenxiang Zheng, Xianhao Ou, Junjia Xiang, Ming Liu, Yujun Zhou, Li Lanyu, Bing Qin

PRDetect: Perturbation-Robust LLM-generated Text Detection Based on Syntax Tree (2025.findings-naacl)

Challenge:	Existing Large Language Models (LLMs) struggle with implicit modality alignment and suboptimal graph linearization.
Approach:	They propose a training-free, test-time adaptive framework that reframes TKG prediction as explicit evidence-driven reasoning.
Outcome:	ExE-LLM outperforms fully trained graph neural networks on four benchmarks . it achieves SOTA performance in inductive settings, significantly outperforming fully trained neural networks .

Copied to clipboard

Xiang Li, Zhiyi Yin, Hexiang Tan, Shaoling Jing, Du Su, Yi Cheng, Huawei Shen, Fei Sun

Syntactic Text Processing

Challenge:	Recent methods for detecting LLM-generated text have shown impressive performance, but in real-world scenarios, users often introduce perturbations to the text.
Approach:	They propose a method that detects syntactic trees that are minimally affected by perturbations and exhibit distinct differences between human-written and LLM-generated text.
Outcome:	The proposed method shows that it is significantly better against perturbations on the HC3 and GPT-3.5-mixed datasets and also has the shortest time expenditure.

Training Language Model to Critique for Better Refinement (2025.findings-acl)

Copied to clipboard

Tianshu Yu, Chao Xiang, Mingchuan Yang, Pei Ke, Bosi Wen, Cunxiang Wang, Jiale Cheng, Li Zhang, Xinyu Mu, Chuxiong Sun, Minlie Huang

A Good Prompt Is Worth Millions of Parameters: Low-resource Prompt-based Learning for Vision-Language Models (2022.acl-long)

Challenge:	Large language models (LLMs) have remarkable evaluation and critique capabilities, providing insightful feedback and identifying flaws in various tasks.
Approach:	They propose a framework to train critic models using refinement signals to generate feedback loops where critiques guide the model in refining its responses.
Outcome:	The proposed framework outperforms traditional methods and open-source models in terms of critique quality and refinement outcomes.

Copied to clipboard

Woojeong Jin, Yu Cheng, Yelong Shen, Weizhu Chen, Xiang Ren

XFormParser: A Simple and Effective Multimodal Multilingual Semi-structured Form Parser (2025.coling-main)

Challenge:	Recent few-shot learning models such as GPT3 are expensive and slow to deploy for real-world applications.
Approach:	They propose a prompt-based low-resource learning method for VL tasks with a few examples . they pre-train a sequence-to-sequence transformer model with prefix and masked language modeling .
Outcome:	The proposed method outperforms Frozen on vision-language tasks with prompt-based learning by 18.2% point.

Copied to clipboard

Xianfu Cheng, Hang Zhang, Jian Yang, Xiang Li, Weixiao Zhou, Fei Liu, Kui Wu, Xiangyuan Guan, Tao Sun, Xianjie Wu, Tongliang Li, Zhoujun Li

Multilinguality

Challenge:	Document AI parsing semi-structured image form is a key information extraction task.
Approach:	They propose a multimodal and multilingual semi-structured FORM PARSER which integrates SER and relation extraction into a unified framework.
Outcome:	The proposed framework achieves up to 1.79% improvement on RE tasks in multilingual and zero-shot settings.

Every Answer Matters: Evaluating Commonsense with Probabilistic Measures (2024.acl-long)

Copied to clipboard

Qi Cheng, Michael Boratko, Pranay Kumar Yelugam, Tim O’Gorman, Nalini Singh, Andrew McCallum, Xiang Li

Semantic Text Processing

Challenge:	Existing commonsense evaluations are often posed as multiple-choice questions, allowing models to exploit systematic biases.
Approach:	They propose a generative task that evaluates common sense via multiple open-ended generations and a method that strongly correlates with human judgments.
Outcome:	The proposed method outperforms strong language model baselines on a dataset of human and machine common sense.

DiffusionDialog: A Diffusion Model for Diverse Dialog Generation with Latent Space (2024.lrec-main)

Copied to clipboard

Jianxiang Xiang, Zhenhua Liu, Haodong Liu, Yin Bai, Jia Cheng, Wenliang Chen

Dialogue Systems & Conversational Agents

Challenge:	Existing studies have tried to introduce discrete or Gaussian-based latent variables to address the one-to-many problem, but the diversity is limited.
Approach:	They propose a diffusion model to enhance the diversity of dialogue generation by using continuous latent variables instead of discrete ones.
Outcome:	The proposed model greatly enhances diversity of dialog response while keeping the coherence.

DIVINE: A Generative Adversarial Imitation Learning Framework for Knowledge Graph Reasoning (D19-1)

Copied to clipboard

Ruiping Li, Xiang Cheng

Knowledge Representation

Challenge:	Existing knowledge graph reasoning methods require numerous trials for path-finding and require meticulous reward engineering to fit specific datasets.
Approach:	They propose a plug-and-play framework that uses generative adversarial imitation learning to enhance existing RL-based methods.
Outcome:	The proposed framework improves existing RL-based methods while eliminating reward engineering.

WordArt Designer: User-Driven Artistic Typography Synthesis using Large Language Models (2023.emnlp-industry)

Copied to clipboard

Jun-Yan He, Zhi-Qi Cheng, Chenyang Li, Jingdong Sun, Wangmeng Xiang, Xianhui Lin, Xiaoyang Kang, Zengke Jin, Yusen Hu, Bin Luo, Yifeng Geng, Xuansong Xie

MarkQA: A large scale KBQA dataset with numerical reasoning (2023.emnlp-main)

Challenge:	Existing typography solutions lack adaptability, creativity, and computational efficiency.
Approach:	They propose a user-driven framework for artistic typography synthesis based on the Large Language Model (LLM) the LLM Engine interprets user inputs and generates actionable prompts for the other modules, transforming abstract concepts into tangible designs.
Outcome:	The proposed framework incorporates four key modules: the LLM Engine, SemTypo, StyTyPo, and TexTyPO.

Copied to clipboard

Xiang Huang, Sitao Cheng, Yuheng Bao, Shanshan Huang, Yuzhong Qu

Reasoning

Challenge:	Existing KBQA datasets are insufficient for numerical reasoning . existing KBqa datasets lack multi-hop reasoning and numerical reasoning.
Approach:	They propose a task that necessitates the ability to perform multi-hop reasoning and numerical reasoning.
Outcome:	The proposed task necessitates the ability to perform multi-hop reasoning and numerical reasoning.

How do autoregressive transformers solve full addition? (2025.emnlp-main)

Copied to clipboard

Wang Peixu, Chen Yu, Yu Ming, Cheng Xiang

Linking Knowledge to Care: Knowledge Graph-Augmented Medical Follow-Up Question Generation (2026.findings-eacl)

Challenge:	Large pre-trained language models have demonstrated impressive capabilities, but there is still much to learn about how they operate.
Approach:	They investigate the ability of the autoregressive transformer to perform basic addition operations by using causal analysis to find that a few different attention heads in the middle layers control the addition carry . they found that due to the lack of global focus on the sequence within these attention heads, the model struggles to handle long-sequence addition tasks.
Outcome:	The model performs basic addition tasks, but it still faces challenges with length generalization.

Copied to clipboard

Liwen Sun, Xiang Yu, Ming Tan, Zhuohao Chen, Anqi Cheng, Ashutosh Joshi, Chenyan Xiong

Mitigating Spurious Correlations via Counterfactual Contrastive Learning (2025.findings-emnlp)

Challenge:	Existing large language models (LLMs) fail to identify information gaps across diverse symptoms.
Approach:	They propose a Knowledge Graph-augmented LLM with active in-context learning to generate relevant and important follow-up questions.
Outcome:	The proposed framework outperforms state-of-the-art methods by 5% - 8% on relevant benchmarks.

Copied to clipboard

Fengxiang Cheng, Chuan Zhou, Xiang Li, Alina Leidinger, Haoxuan Li, Mingming Gong, Fenrong Liu, Robert Van Rooij

Semantic Text Processing

Challenge:	Existing methods to distinguish causally related words from spurious correlations are limited by the number of causally correlated words in a sentence.
Approach:	They propose to use probabilistic probability of necessity and probability of sufficiency to identify causal relationships rather than spurious correlations between words and class labels.
Outcome:	The proposed method is based on a contrastive learning approach name CPNS and is validated on public datasets.

Machine Unlearning of Pre-trained Large Language Models (2024.acl-long)

Copied to clipboard

Jin Yao, Eli Chien, Minxin Du, Xinyao Niu, Tianhao Wang, Zezhou Cheng, Xiang Yue

Will This Idea Spread Beyond Academia? Understanding Knowledge Transfer of Scientific Concepts across Text Corpora (2020.findings-emnlp)

Challenge:	Using curated datasets, we establish a robust benchmark for unlearning performance, demonstrating that these methods are over 105 times more computationally efficient than retraining.
Approach:	They propose a framework for machine unlearning in pre-trained LLMs and integrate gradient ascent with gradient descent on in-distribution data to achieve robustness.
Outcome:	The proposed framework is over 105 times more efficient than retraining on in-distribution data and provides detailed guidelines for efficient hyperparameter tuning in the unlearning process.

Copied to clipboard

Hancheng Cao, Mengjie Cheng, Zhepeng Cen, Daniel McFarland, Xiang Ren

Text Generation

Challenge:	Existing research on knowledge transfer focuses on documents as unit of analysis and follow their transfer into practice for a specific scientific domain.
Approach:	They analyze scientific concepts from corpora and use them to predict knowledge transfer . they find that only a small proportion of these ideas will be used in inventions .
Outcome:	The proposed model predicts the use of scientific concepts in clinical trials and inventions.

OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs (2024.findings-emnlp)

Copied to clipboard

Jintian Zhang, Cheng Peng, Mengshu Sun, Xiang Chen, Lei Liang, Zhiqiang Zhang, Jun Zhou, Huajun Chen, Ningyu Zhang

Green & Sustainable NLP

Challenge:	Recent advances in Large Language Models (LLMs) have significantly enhanced the generative capabilities for various NLP tasks, but they still suffer from hallucinations due to their exclusive reliance on parametric knowledge.
Approach:	They propose a framework that integrates retrieval tokens generated autoregressively into a single LLM to handle both tasks simultaneously in a unified forward pass.
Outcome:	The proposed framework bridges the traditionally separate training approaches for generation and retrieval by incorporating retrieval tokens generated autoregressively.

Rotation Control Unlearning: Quantifying and Controlling Continuous Unlearning for LLM with The Cognitive Rotation Space (2026.findings-acl)

Copied to clipboard

Xiang Zhang, Kun Wei, Xu Yang, Jiahua Li, Su Yan, Cheng Deng

Challenge:	Existing methods to remove undesirable data from Large Language Models suffer from cumulative catastrophic utility loss under continuous unlearning requests.
Approach:	They propose a method that leverages the rotational salience weight of RCU to quantify and control the unlearning degree in the continuous unlearning process.
Outcome:	The proposed method achieves SOTA performance without a retained dataset.

TARGA: Targeted Synthetic Data Generation for Practical Reasoning over Structured Data (2025.acl-long)

Copied to clipboard

Xiang Huang, Jiayu Shen, Shanshan Huang, Sitao Cheng, Xiaxia Wang, Yuzhong Qu

FARSS: Fisher-Optimized Adaptive Low-Rank and Singular-Vector Selection for Knowledge-Preserving Fine-Tuning (2026.findings-acl)

Challenge:	Existing methods for semantic parsing rely on extensive manually annotated datasets and limited generalization capability to unseen examples.
Approach:	They propose a framework that generates high-relevance synthetic data without manual annotation . they generate queries for the queries and use them as demonstrations for in-context learning .
Outcome:	The proposed framework outperforms non-fine-tuned methods on KBQA datasets and shows superior sample efficiency, robustness, and generalization capabilities under non-I.I.D. settings.

Copied to clipboard

Renxing Chen, Ziwei Xiang, Peisong Wang, Hongjian Fang, Meng Li, Fanhu Zeng, Yanan Zhu, Peipei Yang, Xu-Yao Zhang, Jian Cheng

Semantic Text Processing

Challenge:	Low-rank adaptation methods for large language models have limitations in preserving world knowledge and limiting updates to preserve world knowledge.
Approach:	They propose a Fisher-optimized adaptive low Rank and Singular-VectorSelection framework for knowledge-preserving fine-tuning that allows efficient and task-sensitive updates.
Outcome:	The proposed framework outperforms existing methods for knowledge-preserving fine-tuning.

Towards Robust Temporal Activity Localization Learning with Noisy Labels (2024.lrec-main)

Copied to clipboard

Daizong Liu, Xiaoye Qu, Xiang Fang, Jianfeng Dong, Pan Zhou, Guoshun Nan, Keke Tang, Wanlong Fang, Yu Cheng

Robustness in NLP

Challenge:	Existing methods for temporal activity localization are expensive and difficult to satisfy due to subjective labeling.
Approach:	They propose a new TAL setting where a TAL model should be robust to mixed training data with noisy moment boundaries.
Outcome:	The proposed method is significantly more robust to noisy training data than existing methods.

QueryAgent: A Reliable and Efficient Reasoning Framework with Environmental Feedback based Self-Correction (2024.acl-long)

Copied to clipboard

Xiang Huang, Sitao Cheng, Shanshan Huang, Jiayu Shen, Yong Xu, Chaoyun Zhang, Yuzhong Qu

AceGPT, Localizing Large Language Models in Arabic (2024.naacl-long)

Challenge:	Existing methods for semantic parsing fail when hallucinations are encountered . QueryAgent solves a question step-by-step and performs stepwise self-correction .
Approach:	They propose a framework that solves a query step-by-step and performs stepwise self-correction.
Outcome:	The proposed framework outperforms existing methods on GrailQA and GraphQ by 5.7 and 15.0 points.

Copied to clipboard

Huang Huang, Fei Yu, Jianqing Zhu, Xuening Sun, Hao Cheng, Song Dingjie, Zhihong Chen, Mosen Alharthi, Bang An, Juncai He, Ziche Liu, Junying Chen, Jianquan Li, Benyou Wang, Lian Zhang, Ruoyu Sun, Xiang Wan, Haizhou Li, Jinchao Xu

AV-TranSpeech: Audio-Visual Robust Speech-to-Speech Translation (2023.acl-long)

Challenge:	Significant concerns emerge when addressing cultural sensitivity and local values.
Approach:	They propose a localized Large Language Model (LLM) specifically for Arabic, a language imbued with unique cultural characteristics inadequately addressed by current mainstream models.
Outcome:	The proposed model sets the state-of-the-art standard for open Arabic LLMs across various benchmarks.

Copied to clipboard

Rongjie Huang, Huadai Liu, Xize Cheng, Yi Ren, Linjun Li, Zhenhui Ye, Jinzheng He, Lichao Zhang, Jinglin Liu, Xiang Yin, Zhou Zhao

Explain-Analyze-Generate: A Sequential Multi-Agent Collaboration Method for Complex Reasoning (2025.coling-main)

Challenge:	Existing models for speech-to-speech translation suffer from distinct degradation in noisy environments and fail to translate visual speech.
Approach:	They propose a text-based audio-visual speech-to-speech translation model that integrates visual information with audio-only data to improve system robustness.
Outcome:	The proposed model outperforms models trained on audio-only corpus in two languages . it also improves with low-resource audio-visual data, compared with baselines .

Copied to clipboard

WenYuan Gu, JiaLe Han, HaoWen Wang, Xiang Li, Bo Cheng

Beyond Itinerary Planning—A Real-World Benchmark for Multi-Turn and Tool-Using Travel Tasks (2026.acl-long)

Challenge:	Multiagent debate (MAD) is a popular approach for large language models . however, the performance of LLMs is suboptimal in complex reasoning scenarios .
Approach:	They propose a sequential collaboration framework to enable agents to provide constructive assistance to peers by decomposing complex tasks into essential subtasks.
Outcome:	The proposed framework achieves the highest performance on 19 out of 23 tasks and lower costs compared to MAD.

Copied to clipboard

Xiang Cheng, Yulan Hu, Xiangwen Zhang, Lu Xu, Lide Tan, Zheng Pan, Xin Li, Yong Liu

CM-TTS: Enhancing Real Time Text-to-Speech Synthesis Efficiency through Weighted Samplers and Consistency Models (2024.findings-naacl)

Challenge:	Existing studies on LLM performance on travel planning have shown that existing settings are limited due to limited domain coverage, insufficient modeling of users’ implicit preferences in multi-turn conversations, and a lack of evaluation of agents’ capability boundaries.
Approach:	They propose a benchmark to evaluate LLMs' planning and tool-use abilities in real-world settings by collecting user queries, user preferences, and tools from real scenarios.
Outcome:	The proposed benchmark evaluates agents' capabilities in real-world settings and shows that even advanced models exhibit imbalanced performance across different capabilities.

Copied to clipboard

Xiang Li, FanBu FanBu, Ambuj Mehrish, Yingting Li, Jiale Han, Bo Cheng, Soujanya Poria

Identifying while Learning for Document Event Causality Identification (2024.acl-long)

Challenge:	Neural Text-to-Speech systems are a promising approach for high-fidelity speech synthesis . but the efficiency of multi-step sampling in Diffusion Models presents challenges .
Approach:	They propose a novel architecture grounded in consistency models to improve model convergence.
Outcome:	The proposed architecture achieves top-quality speech synthesis in fewer steps without adversarial training or pre-trained model dependencies.

Copied to clipboard

Cheng Liu, Wei Xiang, Bang Wang

Challenge:	Existing studies focus on causality existence, but ignore causal direction.
Approach:	They propose a new identifying while learning mode for the ECI task that takes care of the causal direction and updates events’ representations for boosting next round of causality identification.
Outcome:	The proposed method outperforms the state-of-the-art methods on two public datasets.

Can Large Language Models Act as Ensembler for Multi-GNNs? (2025.emnlp-main)

Copied to clipboard

Hanqi Duan, Yao Cheng, Jianxiang Yu, Yao Liu, Xiang Li

TF-Mamba: Text-enhanced Fusion Mamba with Missing Modalities for Robust Multimodal Sentiment Analysis (2025.findings-emnlp)

Challenge:	Existing graph neural networks lack the inherent semantic understanding capability of rich textual attributes, limiting their effectiveness in applications.
Approach:	They propose a model that integrates multiple GNNs and LLMs to provide an ensemble for multi-GNNs.
Outcome:	The proposed model outperforms existing models in terms of semantic understanding of graph structures and graph structures.

Copied to clipboard

Xiang Li, Xianfu Cheng, Dezhuang Miao, Xiaoming Zhang, Zhoujun Li