Papers by Zhaopeng Tu

ParroT: Translating during Chat using Large Language Models tuned with Human Translation and Feedback (2023.findings-emnlp)

Challenge:	Zero pronouns (ZPs) are often omitted in pro-drop languages, but should be recalled in non-pro-drop language.
Approach:	They propose to analyze the literature on zero pronoun translation after the neural revolution . they uncover that data limitation causes learning bias in languages and domains .
Outcome:	The proposed method and methods are compared to other models and evaluation metrics on different benchmarks.

Copied to clipboard

Wenxiang Jiao, Jen-tse Huang, Wenxuan Wang, Zhiwei He, Tian Liang, Xing Wang, Shuming Shi, Zhaopeng Tu

Draft Model Knows When to Stop: Self-Verification Speculative Decoding for Long-Form Generation (2025.emnlp-main)

Challenge:	Large language models (LLMs) like ChatGPT are only accessible through restricted APIs, which creates barriers to new research and advancements in the field.
Approach:	They propose a framework to enhance and regulate the translation abilities during chat . they reformulate translation data into the instruction-following style and introduce a "Hint" field .
Outcome:	The proposed framework enhances and regulates the translation abilities during chat . it reformulates translation data into the instruction-following style and introduces a "Hint" field .

Copied to clipboard

Ziyin Zhang, Jiahao Xu, Tian Liang, Xingyu Chen, Zhiwei He, Rui Wang, Zhaopeng Tu

Reasoning

Challenge:	Conventional speculative decoding methods use a predefined length policy for proposing drafts, but the reality deviates from this assumption.
Approach:	They propose a self-verification length policy that adaptively determines the lengths of draft sequences by referring to the draft entropy.
Outcome:	The proposed method achieves 17% speedup on MT-Bench and 22% speedup in long-form reasoning.

Learning to Jointly Translate and Predict Dropped Pronouns with a Shared Reconstruction Mechanism (D18-1)

Copied to clipboard

Longyue Wang, Zhaopeng Tu, Andy Way, Qun Liu

Towards Better Modeling Hierarchical Structure for Self-Attention with Ordered Neurons (D19-1)

Challenge:	Pronouns are often omitted in pro-drop languages, such as Chinese . this leads to various translation problems in terms of completeness, syntax and semantics .
Approach:	They propose a reconstruction-based approach to alleviate dropped pronoun (DP) translation problems for neural machine translation models by employing a shared reconstructor and a joint learning approach.
Outcome:	The proposed approach improves translation performance and accuracy of DP predictions.

Copied to clipboard

Jie Hao, Xing Wang, Shuming Shi, Jinfeng Zhang, Zhaopeng Tu

Retrieval-guided Dialogue Response Generation via a Matching-to-Generation Framework (D19-1)

Challenge:	Recent studies have shown that a hybrid of self-attention networks (SANs) and recurrent neural networks (RNNs) outperforms both individual architectures, while not much is known about why the hybrid models work.
Approach:	They propose to use an advanced variant of self-attention networks (SANs) to enhance the strength of hybrid models by introducing a syntax-oriented inductive bias to perform tree-like composition.
Outcome:	The proposed model outperforms both individual models and a standard hybrid model on a machine translation task.

Copied to clipboard

Deng Cai, Yan Wang, Wei Bi, Zhaopeng Tu, Xiaojiang Liu, Shuming Shi

Dialogue Systems & Conversational Agents

Challenge:	generative models for end-to-end sequence generation have been shown promising for this task . however, how to precisely extract a skeleton and how to effectively train a retrieval-guided response generator is still challenging.
Approach:	They propose a framework where skeleton extraction is made by an interpretable matching model and a retrieval-guided response generator is followed by a separate generator.
Outcome:	The proposed framework outperforms baseline models in a variety of experiments.

Progressive Multi-Granularity Training for Non-Autoregressive Translation (2021.findings-acl)

Copied to clipboard

Liang Ding, Longyue Wang, Xuebo Liu, Derek F. Wong, Dacheng Tao, Zhaopeng Tu

How Does Selective Mechanism Improve Self-Attention Networks? (2020.acl-main)

Challenge:	Non-autoregressive translation models are weak at learning high-mode knowledge, argues a new study . despite the improved learning difficulty, there are still complicated word orders and structures in the synthetic sentences, making the NAT performance sub-optimal.
Approach:	They propose to train non-autoregressive translation models to learn fine-grained lower-mode knowledge . they break down sentence-level examples into three types and increase granularities .
Outcome:	The proposed method improves phrase translation accuracy and model reordering ability against strong NAT baselines.

Copied to clipboard

Xinwei Geng, Longyue Wang, Xing Wang, Bing Qin, Ting Liu, Zhaopeng Tu

Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training (2025.acl-long)

Challenge:	Experimental results show that selective SANs outperform the standard SAN by paying more attention to content words that contribute to the meaning of the sentence.
Approach:	They propose to implement selective SANs with a flexible Gumbel-Softmax to improve word order encoding and structure modeling.
Outcome:	The proposed system outperforms the standard SANs on several representative NLP tasks including natural language inference, semantic role labelling, and machine translation.

Copied to clipboard

Youliang Yuan, Wenxiang Jiao, Wenxuan Wang, Jen-tse Huang, Jiahao Xu, Tian Liang, Pinjia He, Zhaopeng Tu

Chain-of-Jailbreak Attack for Image Generation Models via Step by Step Editing (2025.findings-acl)

Challenge:	Large Language Models exhibit a level of intelligence that is both impressive and everevolving, but their ability to refuse generating unsafe content is a double-edged sword.
Approach:	They propose a method to tackle a refusal position bias within safety tuning data that compromises the models’ ability to appropriately refuse generating unsafe content.
Outcome:	The proposed method significantly improves model safety without compromising performance and surpasses baseline methods in defending against attacks.

Copied to clipboard

Wenxuan Wang, Kuiyi Gao, Youliang Yuan, Jen-tse Huang, Qiuzhi Liu, Shuai Wang, Wenxiang Jiao, Zhaopeng Tu

Can’t See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMs (2025.acl-long)

Challenge:	Text-based image generation models, such as Stable Diffusion and DALL-E 3, hold significant potential in content creation and publishing workflows . however, considerable efforts are being made to prevent the generation of harmful content, such abusive, violent, or pornographic material.
Approach:	They propose a chain-of-jailbreak method which decomposes malicious queries into multiple sub-queries and iteratively edits images based on these sub-questions.
Outcome:	The proposed method can bypass safeguards of image generation models for over 60% cases, significantly outperforms other jailbreaking methods (14%)

Copied to clipboard

Wenxuan Wang, Xiaoyuan Liu, Kuiyi Gao, Jen-tse Huang, Youliang Yuan, Pinjia He, Shuai Wang, Zhaopeng Tu

Emotion Classification by Jointly Learning to Lexiconize and Classify (2020.coling-main)

Challenge:	Multimodal Large Language Models (MLLMs) have expanded the capabilities of traditional language models by enabling interaction through both text and images.
Approach:	They propose a multimodal safety awareness benchmark to evaluate MLLMs across 29 safety scenarios with 1,500 carefully curated image-prompt pairs.
Outcome:	The proposed model is able to identify unsafe content and avoid over-sensitivity that can hinder helpfulness.

Copied to clipboard

Deyu Zhou, Shuangzhi Wu, Qing Wang, Jun Xie, Zhaopeng Tu, Mu Li

Text Classification

Challenge:	Existing approaches to identify emotions in short text are limited and lack coverage and inaccuracies when applied to informal short text.
Approach:	They propose a novel emotional network to jointly learn sentence emotions and construct emotion lexicons which are dynamically adapted to a given context.
Outcome:	The proposed model outperforms several approaches proposed in previous studies and achieves new state-of-the-art on the benchmark Twitter dataset.

Modeling Localness for Self-Attention Networks (D18-1)

Copied to clipboard

Baosong Yang, Zhaopeng Tu, Derek F. Wong, Fandong Meng, Lidia S. Chao, Tong Zhang

Modeling Recurrence for Transformer (N19-1)

Challenge:	Existing approaches to model locality for self-attention networks have shown great value for capturing global dependencies.
Approach:	They propose to model localness for self-attention networks to capture local context . they cast localness modeling as a learnable Gaussian bias, which indicates the central and scope of the local region to be paid more attention.
Outcome:	The proposed model improves the ability to capture local context and improves accuracy.

Copied to clipboard

Jie Hao, Xing Wang, Baosong Yang, Longyue Wang, Jinfeng Zhang, Zhaopeng Tu

Multi-Task Learning with Shared Encoder for Non-Autoregressive Machine Translation (2021.naacl-main)

Challenge:	Existing studies show that the lack of recurrence modeling hinders the development of a translation model.
Approach:	They propose to model recurrence for Transformer with an additional recurrent encoder.
Outcome:	The proposed model outperforms the deep model on EnglishGerman and ChineseEnglish translation tasks.

Copied to clipboard

Yongchang Hao, Shilin He, Wenxiang Jiao, Zhaopeng Tu, Michael Lyu, Xing Wang

Self-Attention with Structural Position Representations (D19-1)

Challenge:	Existing non-autoregressive machine translation models have shown significant inference speedup but suffer from inferior translation accuracy.
Approach:	They propose to use AT as an auxiliary task to transfer AT knowledge to NAT models by knowledge distillation.
Outcome:	The proposed method achieves significant improvements over baseline non-Autoregressive machine translation models on WMT14 En-De and WMT16 En-Ro datasets.

Copied to clipboard

Xing Wang, Zhaopeng Tu, Longyue Wang, Shuming Shi

Improving Machine Translation with Human Feedback: An Exploration of Quality Estimation as a Reward Model (2024.naacl-long)

Challenge:	Experimental results show that SANs can't encode positions of input words . SAN's are currently lacking in encoding positions of words based on position-unaware "bagof-words" theory .
Approach:	They propose to augment SANs with structural position representations to capture latent structure of input sentence.
Outcome:	The proposed approach consistently outperforms the sequential representations on translation tasks.

Copied to clipboard

Zhiwei He, Xing Wang, Wenxiang Jiao, Zhuosheng Zhang, Rui Wang, Shuming Shi, Zhaopeng Tu

One Model to Learn Both: Zero Pronoun Prediction and Translation (D19-1)

Challenge:	Existing methods to improve translation quality using human feedback have not been validated.
Approach:	They propose to use quality estimation to predict human preferences for feedback training . they propose to detect incorrect translations and assign a penalty term to the reward scores .
Outcome:	The proposed method outperforms systems using larger parallel corpora by a small amount of monolingual data.

Copied to clipboard

Longyue Wang, Zhaopeng Tu, Xing Wang, Shuming Shi

Unleashing LLM Reasoning Capability via Scalable Question Synthesis from Scratch (2025.acl-long)

Challenge:	Zero pronouns (ZPs) are often omitted in pro-drop languages, but should be recalled in non-pro-drop language.
Approach:	They propose a unified and discourse-aware ZP translation approach for neural MT models . they jointly learn to predict and translate ZPs in an end-to-end manner .
Outcome:	The proposed method improves translation performance and ZP prediction accuracy over baseline models and external models.

Copied to clipboard

Yuyang Ding, Xinyu Shi, Xiaobo Liang, Juntao Li, Zhaopeng Tu, Qiaoming Zhu, Min Zhang

Reasoning

Challenge:	Existing methods to improve the mathematical reasoning capabilities of Large Language Models (LLMs) are limited due to the proprietary nature of the data.
Approach:	They propose a data synthesis method that generates large-scale mathematical reasoning datasets using lightweight 7B-scale models.
Outcome:	The proposed method outperforms existing open-source datasets in both in-domain and out-of-domain evaluations and shows improvements in code reasoning tasks.

Crossing the Reward Bridge: Expanding Reinforcement Learning with Verifiable Rewards Across Diverse Domains (2026.acl-long)

Copied to clipboard

Yi Su, Dian Yu, Linfeng Song, Juntao Li, Haitao Mi, Zhaopeng Tu, Min Zhang, Dong Yu

Challenge:	Reinforcement learning with verifiable rewards (RLVR) has been effective on structured tasks, but its reliance on simple, rule-based verifiers creates a bottleneck.
Approach:	They propose a framework that uses a generative verifier to provide soft, probabilistic rewards.
Outcome:	The proposed framework outperforms existing models up to 10x their size and can be scalable and effective.

Assessing the Ability of Self-Attention Networks to Learn Word Order (P19-1)

Copied to clipboard

Baosong Yang, Longyue Wang, Derek F. Wong, Lidia S. Chao, Zhaopeng Tu

On the Sub-layer Functionalities of Transformer Decoder (2020.findings-emnlp)

Challenge:	Existing studies have attributed SAN to being weak at learning positional information for sequence modeling due to lack of recurrence structure.
Approach:	They propose a word reordering detection task to quantify how well word order information is learned by SAN and RNN.
Outcome:	The proposed task quantifies how well word order information learned by SAN and RNN is learned.

Copied to clipboard

Yilin Yang, Longyue Wang, Shuming Shi, Prasad Tadepalli, Stefan Lee, Zhaopeng Tu

Social Welfare Function Leaderboard: On the Emergence of LLM Agents as the Welfare Dictator (2026.findings-acl)

Challenge:	Existing efforts to interpret the encoder of Transformer-based encoder-decoder architectures for neural machine translation have focused on assessing the encoded representations or interpreting the multi-head self-attentions.
Approach:	They propose to use Transformer-based encoder-decoder architectures to analyze how information is propagated through each module of each decoder layer.
Outcome:	The proposed model can be dropped with minimal loss of performance on three translation datasets and can be used to train and inference faster.

Copied to clipboard

Zhengliang Shi, Ruotian Ma, Jen-tse Huang, Xinbei Ma, Xingyu Chen, Mengru Wang, Qu Yang, Yue Wang, Fanghua Ye, Ziyang Chen, Shanyi Wang, Cixing LI, Wenxuan Wang, Zhaopeng Tu, Xiaolong Li, Zhaochun Ren, Liefeng Bo

EmpDG: Multi-resolution Interactive Empathetic Dialogue Generation (2020.coling-main)

Challenge:	Large language models (LLMs) are increasingly entrusted with high-stakes decisions that affect human welfare.
Approach:	They evaluate 20 state-of-the-art Large language models (LLMs) and 20 LLM dictators to create a social welfare function benchmark.
Outcome:	The proposed model creates dilemma between maximizing collective efficiency and ensuring distributive fairness.

Copied to clipboard

Qintong Li, Hongshen Chen, Zhaochun Ren, Pengjie Ren, Zhaopeng Tu, Zhumin Chen

Dialogue Systems & Conversational Agents

Challenge:	Existing work on empathetic dialogue generation fails to capture the nuances of human emotion and consider the potential of user feedback.
Approach:	They propose a multi-resolution adversarial model - EmpDG - to generate more empathetic responses by exploiting both coarse-grained dialogue-level and fine-grounded token-level emotions.
Outcome:	The proposed model outperforms the state-of-the-art models in both content quality and emotion perceptivity.

Too Good to be Bad: On the Failure of LLMs to Role-Play Villains (2026.findings-acl)

Copied to clipboard

Zihao Yi, Qingxuan Jiang, Ruotian Ma, Xingyu Chen, Qu Yang, Mengru Wang, Fanghua Ye, Ying Shen, Zhaopeng Tu, Xiaolong Li, Liefeng Bo

Challenge:	Large Language Models (LLMs) are increasingly tasked with creative generation, but their ability to portray non-prosocial, antagonistic personas remains largely unexamined.
Approach:	They propose a moral alignment benchmark to test the safety of large language models . they find that models struggle with traits directly antithetical to safety principles .
Outcome:	The proposed model fails to accurately portray morally ambiguous or villainous characters . the model fails most with traits directly antithetical to safety principles .

Adapters for Enhanced Modeling of Multilingual Knowledge and Text (2022.findings-emnlp)

Copied to clipboard

Yifan Hou, Wenxiang Jiao, Meizhen Liu, Carl Allen, Zhaopeng Tu, Mrinmaya Sachan

Dynamic Past and Future for Neural Machine Translation (D19-1)

Challenge:	Large language models learn facts from text corpora, but knowledge graphs contain facts in an explicit triple format, restricting their research and application.
Approach:	They propose to enhance multilingual language models with knowledge from multilingual knowledge graphs . they propose to use cross-lingual entity alignment and facts from MLKGs to improve performance .
Outcome:	The proposed model improves MLLMs with cross-lingual entity alignment and facts from multilingual knowledge graphs for many languages while maintaining performance on other general language tasks.

Copied to clipboard

Zaixiang Zheng, Shujian Huang, Zhaopeng Tu, Xin-Yu Dai, Jiajun Chen

Identifying the Achilles’ Heel: An Iterative Method for Uncovering Factual Errors in Large Language Models (2026.findings-acl)

Challenge:	Neural machine translation models can benefit from modeling translated and untranslated source contents as recurrent states, but this less interpretable recurrence hinders their power to model dynamic updating of and contents during decoding.
Approach:	They propose to model the dynamic updating of and contents during decoding by explicitly separating source words into groups of translated and untranslated contents through parts-to-wholes assignment.
Outcome:	The proposed method achieves significant improvements over both Rnmt and Transformer by producing more adequate translations.

Copied to clipboard

Wenxuan Wang, Yuk-Kit Chan, Zixuan Ling, Shi Juluan, Youliang Yuan, Jen-tse Huang, Yifei Zhang, Wenxiang Jiao, Zhaopeng Tu, Michael R. Lyu

Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atoms (2025.acl-long)

Challenge:	Current methods for evaluating LLMs’ veracity are limited by the need for extensive human labor, test data contamination, or limited scope, hindering efficient and effective exposure of errors.
Approach:	They propose a framework that extracts fact triplets to generate diverse question types using rule-based natural language processing techniques.
Outcome:	The proposed framework can trigger factual errors in up to 55% of questions in large LLMs while maintaining coverage of questions.

Copied to clipboard

Mengru Wang, Ziwen Xu, Shengyu Mao, Shumin Deng, Zhaopeng Tu, Huajun Chen, Ningyu Zhang

Salute the Classic: Revisiting Challenges of Machine Translation in the Age of Large Language Models (2025.tacl-1)

Challenge:	Recent research has explored the use of sparse autoencoders (SAE) to disentangle knowledge in high-dimensional spaces for steering.
Approach:	They propose a method that isolates and manipulates disentangled knowledge components to enhance safety by using sparse autoencoders to disentangle knowledge in high-dimensional spaces for steering.
Outcome:	The proposed method is able to isolate and manipulate disentangled knowledge components to enhance safety in large reasoning models.

Copied to clipboard

Jianhui Pang, Fanghua Ye, Derek Fai Wong, Dian Yu, Shuming Shi, Zhaopeng Tu, Longyue Wang

Sentient Agent as a Judge: Evaluating Higher-Order Social Cognition in Large Language Models (2026.findings-acl)

Challenge:	a recent study revisits six core challenges that have influenced the evolution of Neural Machine Translation (NMT) domain mismatch, amount of parallel data, rare word prediction, translation of long sentences and sub-optimal beam search remain challenges in LLMs.
Approach:	They revisit core challenges that have acted as benchmarks for progress in NMT . they propose to revisit these challenges and offer insights into their relevance .
Outcome:	The proposed models significantly improve translation of sentences containing approximately 80 words, even translating documents up to 512 words.

Copied to clipboard

Bang Zhang, Ruotian Ma, Qingxuan Jiang, Peisong Wang, Jiaqi Chen, Zheng Xie, Xingyu Chen, Yue Wang, Fanghua Ye, Jian Li, Yifan Yang, Zhaopeng Tu, Xiaolong Li

DRPruning: Efficient Large Language Model Pruning through Distributionally Robust Optimization (2025.acl-long)

Challenge:	Large language models (LLMs) have evolved from statistical sequence predictors to sophisticated autonomous agents capable of reasoning, planning, and sustaining multi-turn conversa-tions.
Approach:	They propose a system that instantiates a "Sentient Agent" that simulates human-like emotional changes and inner thoughts to provide a more realistic evaluation of the model in multi-turn conversations.
Outcome:	The proposed framework measures the agent's higher-order social cognition in multi-turn conversations.

Copied to clipboard

Hexuan Deng, Wenxiang Jiao, Xuebo Liu, Jing Li, Min Zhang, Zhaopeng Tu

Benchmarking and Improving Long-Text Translation with Large Language Models (2024.findings-acl)

Challenge:	Structured pruning reduces model size but often causes uneven degradation across domains, leading to biased performance.
Approach:	They propose a method that dynamically adjusts the data distribution during training to restore balanced performance across heterogeneous and multi-tasking data.
Outcome:	Experiments in monolingual and multilingual settings show that the proposed method surpasses similarly sized models in pruning and continued pretraining over perplexity, downstream tasks, and instruction tuning.

Copied to clipboard

Longyue Wang, Zefeng Du, Wenxiang Jiao, Chenyang Lyu, Jianhui Pang, Leyang Cui, Kaiqiang Song, Derek Wong, Shuming Shi, Zhaopeng Tu

Document-Level Machine Translation with Large Language Models (2023.emnlp-main)

Challenge:	Recent studies have illuminated the promising capabilities of large language models (LLMs) in handling long texts.
Approach:	They construct a benchmark dataset specifically designed for the finetuning and evaluation of large language models (LLMs) they compare LLMs with MT models and find they exhibit shortcomings in long-text domains .
Outcome:	The proposed model performs better in long-text translation, and its performance diminishes as document size increases.

Copied to clipboard

Longyue Wang, Chenyang Lyu, Tianbo Ji, Zhirui Zhang, Dian Yu, Shuming Shi, Zhaopeng Tu

Data Rejuvenation: Exploiting Inactive Training Examples for Neural Machine Translation (2020.emnlp-main)

Challenge:	Large language models (LLMs) such as ChatGPT can produce coherent, cohesive, relevant, and fluent answers for various natural language processing tasks.
Approach:	They examine the impact of different prompts on document-level translation quality and discourse phenomena using figures and lines, which are invisible to GPT-4.
Outcome:	The proposed models outperform commercial MT systems and advanced document-level MT methods on a number of benchmarks and show potential to become a new paradigm for document- level translation.

Copied to clipboard

Wenxiang Jiao, Xing Wang, Shilin He, Irwin King, Michael Lyu, Zhaopeng Tu

Bridging the Data Gap between Training and Inference for Unsupervised Neural Machine Translation (2022.acl-long)

Challenge:	Large-scale training datasets make training neural machine translation models difficult.
Approach:	They propose to identify inactive training examples which contribute less to the model performance and introduce data rejuvenation to improve NMT models' training.
Outcome:	The proposed framework stabilizes and accelerates the training process of NMT models, resulting in models with better generalization capability.

Copied to clipboard

Zhiwei He, Xing Wang, Rui Wang, Shuming Shi, Zhaopeng Tu

Don’t Get Lost in the Trees: Streamlining LLM Reasoning by Overcoming Tree Search Exploration Pitfalls (2025.acl-long)

Challenge:	Experimental results show that backtranslation improves UNMT performance by reducing the data gap between training and inference.
Approach:	They propose an online method to remedy the source discrepancy between training and inference . they use pseudo parallel data with translated source and translated target to mimic inference scenario .
Outcome:	The proposed method outperforms baselines on several widely-used language pairs by remedying the style and content gaps.

Copied to clipboard

Ante Wang, Linfeng Song, Ye Tian, Dian Yu, Haitao Mi, Xiangyu Duan, Zhaopeng Tu, Jinsong Su, Dong Yu

Green & Sustainable NLP

Challenge:	Recent advances in tree search algorithms guided by verifiers have significantly enhanced the reasoning capabilities of large language models (LLMs), but at the cost of increased computational resources.
Approach:	They propose an e ffici ent tree sear ch framework that is a plug-and-play system compatible with various tree search algorithms.
Outcome:	The proposed framework reduces computational costs and prioritizes resource allocation to harder tasks (Levels 3-4) over simpler ones (Level 1-2), addressing both over-exploration in basic problems and under-exploation in complex cases.

Information Aggregation for Multi-Head Attention with Routing-by-Agreement (N19-1)

Copied to clipboard

Jian Li, Baosong Yang, Zi-Yi Dou, Xing Wang, Michael R. Lyu, Zhaopeng Tu

GuoFeng: A Benchmark for Zero Pronoun Recovery and Translation (2022.emnlp-main)

Challenge:	Existing studies focus on extracting informative or distinct partial-representations from different subspaces, while few studies have paid attention to the aggregation of the extracted partial-Representations.
Approach:	They propose to use a routing-by-agreement algorithm to improve multi-head attention by iteratively updating the proportion of how much a part should be assigned to a whole based on agreement between parts and wholes.
Outcome:	The proposed algorithm improves the information aggregation for multi-head attention over the standard linear transformation on linguistic probing and machine translation tasks.

Copied to clipboard

Mingzhou Xu, Longyue Wang, Derek F. Wong, Hongye Liu, Linfeng Song, Lidia S. Chao, Shuming Shi, Zhaopeng Tu

On the Copying Behaviors of Pre-Training for Neural Machine Translation (2021.findings-acl)

Challenge:	ZPs are often omitted when they can be pragmatically or grammatically inferred from intraand inter-sentential contexts.
Approach:	They propose a benchmark testset for target evaluation on Chinese-English ZP translation.
Outcome:	The proposed testset covers five genres and identifies current challenges for evaluation.

Copied to clipboard

Xuebo Liu, Longyue Wang, Derek F. Wong, Liang Ding, Lidia S. Chao, Shuming Shi, Zhaopeng Tu

Rethinking the Value of Transformer Components (2020.coling-main)

Challenge:	Existing studies show that initializing NMT models with pre-trained language models (LM) can speed up the model training and boost the model performance.
Approach:	They propose a method to control copying behaviors in NMT models by initializing them with pre-trained language models (LM) they propose to use a metric called copy ratio to control the copying behavior in decoding.
Outcome:	The proposed method improves translation performance by controlling copying behaviors for pre-training based models.

Copied to clipboard

Wenxuan Wang, Zhaopeng Tu

Redistributing Low-Frequency Words: Making the Most of Monolingual Data in Non-Autoregressive Translation (2022.acl-long)

Challenge:	Empirical results show that certain components are more important than others . we propose a new training strategy that can improve Transformer models by distinguishing unimportant components .
Approach:	They propose a training strategy that distinguishes the unimportant components in training . they compare the impact of individual component (sub-layer) on model performance .
Outcome:	The proposed training strategy can improve translation performance by distinguishing unimportant components in training.

Copied to clipboard

Liang Ding, Longyue Wang, Shuming Shi, Dacheng Tao, Zhaopeng Tu

Multi-Head Attention with Disagreement Regularization (D18-1)

Challenge:	Knowledge distillation (KD) is the preliminary step for training non-autoregressive translation models, but it can lose important information for translating low-frequency words.
Approach:	They propose a knowledge distillation method which trains NAT student on external monolingual data with AT teacher trained on the original bilingual data.
Outcome:	Extensive experiments on eight WMT benchmarks show that monolingual KD outperforms the standard KD by improving low-frequency word translation without introducing any computational cost.

Copied to clipboard

Jian Li, Zhaopeng Tu, Baosong Yang, Michael R. Lyu, Tong Zhang

Scaling Back-Translation with Domain Text Generation for Sign Language Gloss Translation (2023.eacl-main)

Challenge:	Existing methods to encourage diversity among multi-head attention are limited.
Approach:	They propose a disagreement regularization term to encourage diversity among attention heads . they validated their approach on EnglishGerman and ChineseEnglish translation tasks .
Outcome:	The proposed approach improves translation performance across language pairs on English-German and Chinese-English translation tasks.

Copied to clipboard

Jinhui Ye, Wenxiang Jiao, Xing Wang, Zhaopeng Tu

On the Complementarity between Pre-Training and Back-Translation for Neural Machine Translation (2021.findings-emnlp)

Challenge:	Sign language gloss translation aims to translate the sign glosses into spoken language texts, which is challenging due to the scarcity of labeled gloss-text parallel data.
Approach:	They propose a back translation technique that generates pseudo-parallel data by translating in-domain spoken language texts into sign glosses.
Outcome:	The proposed method outperforms the BT methods on three benchmarks of sign language gloss translation in different languages.

Copied to clipboard

Xuebo Liu, Longyue Wang, Derek F. Wong, Liang Ding, Lidia S. Chao, Shuming Shi, Zhaopeng Tu

Cross-modality Data Augmentation for End-to-End Sign Language Translation (2023.findings-emnlp)

Challenge:	Experimental results show that PT and BT are nicely complementary to each other.
Approach:	They introduce two probing tasks for PT and BT respectively and investigate their complementarity.
Outcome:	The proposed methods establish state-of-the-art on the WMT16 English-Romanian and English-Russian benchmarks.

Copied to clipboard

Jinhui Ye, Wenxiang Jiao, Xing Wang, Zhaopeng Tu, Hui Xiong

Revisiting the Markov Property for Machine Translation (2024.findings-eacl)

Challenge:	End-to-end sign language translation (SLT) aims to convert sign language videos into spoken language texts without intermediate representations.
Approach:	They propose a cross-modality data-augmented framework to transfer gloss-to-text translation capabilities to end-to end sign language translation.
Outcome:	The proposed framework outperforms baseline models on two widely used SLT datasets.

Copied to clipboard

Cunxiao Du, Hao Zhou, Zhaopeng Tu, Jing Jiang

Not All Countries Celebrate Thanksgiving: On the Cultural Dominance in Large Language Models (2024.acl-long)

Challenge:	Statistical machine translation (SMT) has employed Markov models, but autoregressive models are less effective.
Approach:	They propose to use a Markov Autoregressive Transformer to model neural machine translation using four WMT benchmarks.
Outcome:	The proposed model performs better than autoregressive models on four WMT benchmarks.

Copied to clipboard

Wenxuan Wang, Wenxiang Jiao, Jingyuan Huang, Ruyi Dai, Jen-tse Huang, Zhaopeng Tu, Michael Lyu

Towards Understanding Neural Machine Translation with Word Importance (D19-1)

Challenge:	e.g., ChatGPT often provides inappropriate English-culture-related answers when users ask in non-English languages.
Approach:	They build a benchmark of concrete and abstract cultural objects to evaluate the cultural dominance issue in large language models.
Outcome:	The proposed model can significantly mitigate cultural dominance issue in large language models . the model can provide accurate answers in English, while the model is ethically sound .

Copied to clipboard

Shilin He, Zhaopeng Tu, Xing Wang, Longyue Wang, Michael Lyu, Shuming Shi

Skeleton-to-Response: Dialogue Generation Guided by Retrieval Memory (N19-1)

Challenge:	Neural machine translation (NMT) has advanced the state-of-the-art on various language pairs, but the interpretability of NMT remains unsatisfactory.
Approach:	They propose to attribute NMT output to every input word using a gradient-based method to measure word importance.
Outcome:	The proposed method is superior on identifying input words with higher influence on translation performance.

Copied to clipboard

Deng Cai, Yan Wang, Wei Bi, Zhaopeng Tu, Xiaojiang Liu, Wai Lam, Shuming Shi

Dialogue Systems & Conversational Agents

Challenge:	Existing generative dialogue models generate responses from input queries . however, the results are limited and the models are unsatisfactory .
Approach:	They propose a framework which exploits retrieval results via a skeleton-to-response paradigm . they extract a query skelet and use it to generate a new skele and response .
Outcome:	The proposed approach significantly improves the informativeness of the generated responses.

Exploiting Sentential Context for Neural Machine Translation (P19-1)

Copied to clipboard

Xing Wang, Zhaopeng Tu, Longyue Wang, Shuming Shi

Rejuvenating Low-Frequency Words: Making the Most of Parallel Data in Non-Autoregressive Translation (2021.acl-long)

Challenge:	Existing approaches to exploit sentential context for machine translation are not well studied.
Approach:	They propose a shallow sentential context that exploits top encoder layer, and a deep sentential one that aggregates sentential representations from all internal layers.
Outcome:	The proposed model outperforms the strong Transformer model on the English-German and English-French benchmarks.

Copied to clipboard

Liang Ding, Longyue Wang, Xuebo Liu, Derek F. Wong, Dacheng Tao, Zhaopeng Tu

ngram-OAXE: Phrase-Based Order-Agnostic Cross Entropy for Non-Autoregressive Machine Translation (2022.coling-1)

Challenge:	Knowledge distillation (KD) is commonly used to construct synthetic data for training non-autoregressive translation models.
Approach:	They propose to use knowledge distillation to generate training data for non-autoregressive translation models by leveraging pretraining.
Outcome:	The proposed approach achieves 28.2 and 33.9 BLEU points on the WMT14 English-German and WMT16 Romanian-English datasets.

Copied to clipboard

Cunxiao Du, Zhaopeng Tu, Longyue Wang, Jing Jiang

CogDual: Enhancing Dual Cognition of LLMs via Reinforcement Learning with Implicit Rule-Based Rewards (2025.emnlp-main)

Challenge:	Recent studies have incorporated approaches to improving the standard cross-entropy loss to ameliorate the effect of multimodality.
Approach:	They propose a new training oaxe loss which removes the penalty of word order errors in the standard cross-entropy loss.
Outcome:	Extensive experiments on NAT benchmarks show that the proposed approach improves translation quality and improves model performance.

Copied to clipboard

Cheng Liu, Yifei Lu, Fanghua Ye, Jian Li, Xingyu Chen, Feiliang Ren, Zhaopeng Tu, Xiaolong Li

Challenge:	Existing approaches to role-playing language models rely on prompt engineering or supervised fine-tuning to emulate character behaviors but neglect the underlying cognitive mechanisms driving these behaviors.
Approach:	They propose a novel RPLA adopting a cognize-then-respond reasoning paradigm that leverages dual cognition for more contextually grounded and psychologically coherent responses.
Outcome:	The proposed RPLA outperforms baselines and generalizes effectively across diverse role-playing tasks.

On the Inference Calibration of Neural Machine Translation (2020.acl-main)

Copied to clipboard

Shuo Wang, Zhaopeng Tu, Shuming Shi, Yang Liu

On the Sparsity of Neural Machine Translation Models (2020.emnlp-main)

Challenge:	Existing studies show that NMT models trained with label smoothing are well-calibrated on ground-truth training data, but miscalibration remains a challenge during inference due to the discrepancy between training and inference.
Approach:	They propose a graduated label smoothing method that can improve inference calibration and translation performance.
Outcome:	The proposed method improves both inference calibration and translation performance.

Copied to clipboard

Yong Wang, Longyue Wang, Victor Li, Zhaopeng Tu

RAST: Domain-Robust Dialogue Rewriting as Sequence Tagging (2021.emnlp-main)

Challenge:	Modern neural machine translation models employ a large number of parameters, which leads to serious over-parameterization.
Approach:	They propose to prune parameters to improve the model by +0.8 BLEU points and to reallocate them to enhance the ability of modeling low-level lexical information.
Outcome:	The pruned parameters improve the model by +0.8 BLEU points and the rejuvenated parameters enhance the ability to model low-level lexical information.

Copied to clipboard

Jie Hao, Linfeng Song, Liwei Wang, Kun Xu, Zhaopeng Tu, Dong Yu

Robustness in NLP

Challenge:	Existing models for dialogue rewriting suffer from the robustness issue, i.e., performances drop dramatically when testing on a different dataset.
Approach:	They propose a sequence-tagging-based approach that reduces the search space while preserving the core of the task.
Outcome:	The proposed model significantly reduces the search space while still covering the core of the task.

On the Language Coverage Bias for Neural Machine Translation (2021.findings-acl)

Copied to clipboard

Shuo Wang, Zhaopeng Tu, Zhixing Tan, Shuming Shi, Maosong Sun, Yang Liu

Unsupervised Sign Language Translation and Generation (2024.findings-acl)

Challenge:	Language coverage bias is important for neural machine translation because of the target-original training data.
Approach:	They propose two approaches to alleviate the language coverage bias problem by explicitly distinguishing between the source-and target-original training data.
Outcome:	The proposed methods improve translation tasks on both back-and forward-translation and their tagged variants.

Copied to clipboard

Zhengsheng Guo, Zhiwei He, Wenxiang Jiao, Xing Wang, Rui Wang, Kehai Chen, Zhaopeng Tu, Yong Xu, Min Zhang

Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models (2024.acl-long)

Challenge:	Experimental results on the BBC-Oxford Sign Language dataset reveal that USLNet achieves competitive results compared to supervised baseline models.
Approach:	They propose an unsupervised sign language translation and generation network that learns from abundant single-modality data without parallel sign language data.
Outcome:	The proposed model achieves competitive results compared to baseline models on the BBC-Oxford Sign Language dataset and Open-Domain American Sign Language data.

Copied to clipboard

Zhiwei He, Binglin Zhou, Hongkun Hao, Aiwei Liu, Xing Wang, Zhaopeng Tu, Zhuosheng Zhang, Rui Wang

Exploiting Deep Representations for Neural Machine Translation (D18-1)

Challenge:	Existing text watermarking technologies lack consistency when texts are translated into different languages.
Approach:	They propose a cross-lingual watermark removal attack to bypass watermarking by first obtaining a response from an LLM in a pivot language and then translating it into the target language.
Outcome:	The proposed method can remove watermarks without performance loss by obtaining a response from an LLM in a pivot language and then translating it into the target language.

Copied to clipboard

Zi-Yi Dou, Zhaopeng Tu, Xing Wang, Shuming Shi, Tong Zhang

Self-Training Sampling with Monolingual Data Uncertainty for Neural Machine Translation (2021.acl-long)

Challenge:	Neural machine translation models typically implement encoder and decoder as multiple layers, but only the top layers are leveraged in the subsequent process, which misses the opportunity to exploit useful information embedded in other layers.
Approach:	They propose to expose all of these signals with layer aggregation and multi-layer attention mechanisms and introduce an auxiliary regularization term to encourage different layers to capture diverse information.
Outcome:	The proposed approach exposes all of these signals with layer aggregation and multi-layer attention mechanisms on widely-used translation datasets.

Copied to clipboard

Wenxiang Jiao, Xing Wang, Zhaopeng Tu, Shuming Shi, Michael Lyu, Irwin King

Understanding and Improving Sequence-to-Sequence Pretraining for Neural Machine Translation (2022.acl-long)

Challenge:	Experimental results show that enhancing the learning on uncertain monolingual sentences improves the translation quality of high-uncertainty sentences and also benefits the prediction of low-frequency words at the target side.
Approach:	They propose to use monolingual data to augment model training with synthetic parallel data by selecting the most informative monolingual sentences to complement the parallel data.
Outcome:	The proposed approach improves the performance of natural language models by selecting the most informative monolingual sentences.

Copied to clipboard

Wenxuan Wang, Wenxiang Jiao, Yongchang Hao, Xing Wang, Shuming Shi, Zhaopeng Tu, Michael Lyu

Convolutional Self-Attention Networks (N19-1)

Challenge:	Existing studies on self-supervised pretraining for machine translation have focused on the jointly pretrained decoder .
Approach:	They propose a method to improve neural machine translation by jointly pretrained decoder . they propose two strategies to remedy the domain and objective discrepancies .
Outcome:	The proposed approach improves translation performance and model robustness on three language pairs.

Copied to clipboard

Baosong Yang, Longyue Wang, Derek F. Wong, Lidia S. Chao, Zhaopeng Tu

Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate (2024.emnlp-main)

Challenge:	Existing models of self-attention networks lack the ability to capture dependencies regardless of distance and can be enhanced with multi-head attention.
Approach:	They propose a convolutional self-attention network which can be enhanced by multi-head attention by allowing the model to attend to information from different representation subspaces.
Outcome:	The proposed model outperforms existing models on improving locality of SANs on different language pairs and model settings.

Copied to clipboard

Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shuming Shi, Zhaopeng Tu

Towards Robust Neural Machine Translation (P18-1)

Challenge:	Modern large language models (LLMs) have shown remarkable performance on general language tasks but struggle on complex reasoning tasks.
Approach:	They propose a multi-agent debate framework that encourages divergent thinking in LLMs . they propose to break debate and use a judge to obtain a final solution .
Outcome:	The proposed framework encourages divergent thinking in large language models . it is able to generate novel thoughts even if initial stance is incorrect .

Copied to clipboard

Yong Cheng, Zhaopeng Tu, Fandong Meng, Junjie Zhai, Yang Liu

Multi-Granularity Self-Attention for Neural Machine Translation (D19-1)

Challenge:	Small perturbations in the input can severely distort intermediate representations and thus impact translation quality of neural machine translation models.
Approach:	They propose adversarial stability training to make encoder and decoder robust to perturbations by enabling them to behave similarly for the original input and its perturbed counterpart.
Outcome:	The proposed approach improves translation quality and robustness over strong models on Chinese-English, English-German and English-French translation tasks.

Copied to clipboard

Jie Hao, Xing Wang, Shuming Shi, Jinfeng Zhang, Zhaopeng Tu

A Template-based Method for Constrained Neural Machine Translation (2022.emnlp-main)

Challenge:	Existing neural machine translation models use a deep multi-head self-attention network with no explicit phrase information.
Approach:	They propose a neural network that combines multi-head self-attention and phrase modeling to train attention heads to attend to phrases in either n-gram or syntactic formalisms.
Outcome:	The proposed approach improves on English-to-German and NIST Chinese-to English translation tasks.

Copied to clipboard

Shuo Wang, Peng Li, Zhixing Tan, Zhaopeng Tu, Maosong Sun, Yang Liu

Context-Aware Cross-Attention for Non-Autoregressive Translation (2020.coling-main)

Challenge:	Existing methods to solve this problem can not satisfy the following three desiderata: (1) high translation quality, (2) high match accuracy, and (3) low latency.
Approach:	They propose a template-based method that can provide high translation quality and match accuracy and a low latency inference.
Outcome:	The proposed method outperforms baselines in lexically and structurally constrained translation tasks and can be used in a variety of applications.

Copied to clipboard

Liang Ding, Longyue Wang, Di Wu, Dacheng Tao, Zhaopeng Tu

Addressing Entity Translation Problem via Translation Difficulty and Context Diversity (2024.findings-acl)

Challenge:	Existing studies have shown that non-autoregressive translation models can predict all tokens independently and simultaneously.
Approach:	They propose to enhance signals of neighbour source tokens into conventional cross-attention to address a locality perception problem in NAT cross- attention.
Outcome:	The proposed approach improves translation quality over strong NAT baselines on representative datasets.

Copied to clipboard

Tian Liang, Xing Wang, Mingming Yang, Yujiu Yang, Shuming Shi, Zhaopeng Tu

All Languages Matter: On the Multilingual Safety of LLMs (2024.findings-acl)

Challenge:	Neural machine translation systems often produce inadequate translations for named entities.
Approach:	They propose a data augmentation strategy to enhance the accuracy of named entity translation by retraining the target named entity pair.
Outcome:	The proposed method improves translation accuracy across test sets and terminology tests.

Copied to clipboard

Wenxuan Wang, Zhaopeng Tu, Chang Chen, Youliang Yuan, Jen-tse Huang, Wenxiang Jiao, Michael Lyu

BatonVoice: An Operationalist Framework for Enhancing Controllable Speech Synthesis with Linguistic Intelligence from LLMs (2026.acl-long)

Challenge:	Existing safety benchmarks only concern the safety in one language, e.g. the majority language in the pretraining data such as English.
Approach:	They propose a prompting method to improve multilingual safety of ChatGPT by enhancing cross-lingual generalization of safety alignment.
Outcome:	The proposed method can significantly reduce the ratio of unsafe responses by 42% for non-English queries.

Copied to clipboard

Yue Wang, Ruotian Ma, Xingyu Chen, Zhengliang Shi, Morunliu Yang, Wanshun Chen, Huang Liu, Jiadi Yao, Xin He, Qu Yang, Qingxuan Jiang, Fanghua Ye, Juntao Li, Zhaopeng Tu, Xiaolong Li, Liefeng Bo, Min Zhang

Speech & Audio in NLP

Challenge:	Existing approaches often fail to leverage the linguistic intelligence of Large Language Models (LLMs) Existing models lack the ability to follow text instructions for controllable Text-to-Speech (TTS).
Approach:	They propose a framework where an LLM acts as a conductor, understanding user instructions and generating a textual plan - explicit vocal features.
Outcome:	The proposed model outperforms open- and closed-source models in speech synthesis and achieves zero-shot cross-lingual generalization.

Target Foresight Based Attention for Neural Machine Translation (N18-1)

Copied to clipboard

Xintong Li, Lemao Liu, Zhaopeng Tu, Shuming Shi, Max Meng

Revisiting Non-Autoregressive Translation at Scale (2023.findings-acl)

Challenge:	Empirical experiments on Chinese-to-English and Japanese-to English datasets show that the proposed attention model delivers significant improvements in terms of alignment error rate and BLEU.
Approach:	They propose to explicitly access the target foresight word in the attention model to improve alignment and translation accuracy.
Outcome:	Empirical results show that the proposed model improves alignment error rate and BLEU on Chinese-to-English and Japanese-toEnglish datasets.

Copied to clipboard

Zhihao Wang, Longyue Wang, Jinsong Su, Junfeng Yao, Zhaopeng Tu