Papers by Hongfei Xu

Just Like a Human Would, Direct Access to Sarcasm Augmented with Potential Result and Reaction (2023.acl-long)

Challenge:	Patronizing and condescending language is an essential branch of toxic language . pre-trained language models perform poorly in detecting PCL due to its implicit toxicity traits .
Approach:	They propose a novel LLM benchmark for patronizing and condescending language . they use a dataset to analyze the toxicity of patronizing condescending languages .
Outcome:	The proposed model can detect patronizing and condescending language (PCL) the model can be used to analyze the toxicity of the language and to improve the detection.

Copied to clipboard

Changrong Min, Ximing Li, Liang Yang, Zhilin Wang, Bo Xu, Hongfei Lin

Lipschitz Constrained Parameter Initialization for Deep Transformers (2020.acl-main)

Challenge:	sarcasm is a form of irony conveying mockery and contempt . social media has become increasingly popular for identifying sarcasm .
Approach:	They develop a method to detect sarcasm from social media using augmented potentials.
Outcome:	The proposed method outperforms baselines on benchmark datasets.

Copied to clipboard

Hongfei Xu, Qiuhui Liu, Josef van Genabith, Deyi Xiong, Jingyi Zhang

Learning Hard Retrieval Decoder Attention for Transformers (2021.findings-emnlp)

Challenge:	Existing studies show that deep Transformers have difficulty in training even with residual connection and layer normalization.
Approach:	They propose a method that leverages the Lipschitz constraint on the initialization of Transformer parameters to ease the optimization difficulties caused by its multi-layer encoder/decoder structure.
Outcome:	The proposed model outperforms previous RNN/CNN models but fails to converge with the original computation order.

Copied to clipboard

Hongfei Xu, Qiuhui Liu, Josef van Genabith, Deyi Xiong

Is LLM an Overconfident Judge? Unveiling the Capabilities of LLMs in Detecting Offensive Language with Annotation Disagreement (2025.findings-acl)

Challenge:	In this paper, we show that learning a hard retrieval attention that attends to a single token in a sentence is 1.43 times faster than the standard scaled dot-product attention.
Approach:	They propose a method to learn hard retrieval attention where an attention head attends to a single token in a sentence rather than all tokens.
Outcome:	The proposed method is 1.43 times faster in decoding while preserving translation quality on a wide range of MT tasks.

Copied to clipboard

Junyu Lu, Kai Ma, Kaichun Wang, Kelaiti Xiao, Roy Ka-Wei Lee, Bo Xu, Liang Yang, Hongfei Lin

Responsible & Trustworthy NLP

Challenge:	Existing studies focus on evaluating large language models' ability to handle disagreement cases.
Approach:	They evaluate the performance of large language models in detecting offensive language at varying levels of agreement.
Outcome:	The proposed model improves detection accuracy and model alignment with human judgment by using disagreement samples in training.

Learning Source Phrase Representations for Neural Machine Translation (2020.acl-main)

Copied to clipboard

Hongfei Xu, Josef van Genabith, Deyi Xiong, Qiuhui Liu, Jingyi Zhang

Dual-teacher Knowledge Distillation for Low-frequency Word Translation (2024.findings-emnlp)

Challenge:	Existing approaches to machine translation have been shown to be effective for long sentences . however, the attentional network can't capture long-distance dependencies .
Approach:	They propose a multi-head attention mechanism which generates phrase representations from token representations and incorporates them into the Transformer translation model to enhance its ability to capture long-distance relationships.
Outcome:	The proposed model can be computed in parallel and improves on the WMT 14 tasks.

Copied to clipboard

Yifan Guo, Hongying Zan, Hongfei Xu

A Self-Distillation Recipe for Neural Machine Translation (2025.findings-acl)

Challenge:	Neural machine translation models are trained on parallel corpora with unbalanced word frequency distribution, resulting in high-frequency words being ignored.
Approach:	They propose to employ a low-frequency teacher model that excels in translating low- frequency words to guide the learning of the student model.
Outcome:	The proposed method achieves +0.64 BLEU improvements over the state-of-the-art method on the low-frequency translation task while maintaining the translation quality of high-frequency words.

Copied to clipboard

Hongfei Xu, Zhuofei Liang, Qiuhui Liu, Lingling Mu

RealMedDial: A Real Telemedical Dialogue Dataset Collected from Online Chinese Short-Video Clips (2022.coling-1)

Challenge:	Existing methods for Neural Machine Translation (NMT) have been proven effective in improving the performance of computer vision tasks without pre-training a teacher.
Approach:	They propose a rank-order augmented Pearson correlation loss and an iterative distillation method to prevent the discrepancy of predictions between the student and a stronger teacher from disturbing the training.
Outcome:	The proposed method can lead to significant improvements over the strong Transformer baseline on low/middle/high-resource tasks, obtaining comparable or better performance with fewer layers.

Copied to clipboard

Bo Xu, Hongtong Zhang, Jian Wang, Xiaokun Zhang, Dezhi Hao, Linlin Zong, Hongfei Lin, Fenglong Ma

Dialogue Systems & Conversational Agents

Challenge:	Existing medical dialogue systems are limited by the lack of corpora and data from real scenarios.
Approach:	They construct a Chinese medical dialogue dataset based on real medical consultations.
Outcome:	The proposed dataset is applicable to a wide range of NLP tasks with respect to medical dialogue.

Prototype Tuning: A Meta-Learning Approach for Few-Shot Document-Level Relation Extraction with Large Language Models (2025.findings-naacl)

Copied to clipboard

Dinghao Pan, Yuanyuan Sun, Bo Xu, Jiru Li, Zhihao Yang, Ling Luo, Hongfei Lin, Jian Wang

Dynamically Adjusting Transformer Batch Size by Monitoring Gradient Direction Change (2020.acl-main)

Challenge:	Few-Shot Document-Level Relation Extraction (FSDLRE) aims to develop models capable of generalizing to new categories with minimal support examples.
Approach:	They propose a meta-training approach to train Large Language Models to improve their ICL capabilities . they construct simulated episodes using relation types that do not overlap with test corpus .
Outcome:	Experimental results show that the proposed approach outperforms baseline models on few-shot tasks.

Copied to clipboard

Hongfei Xu, Josef van Genabith, Deyi Xiong, Qiuhui Liu

HyperHatePrompt: A Hypergraph-based Prompting Fusion Model for Multimodal Hate Detection (2025.coling-main)

Challenge:	Compared to previous studies, the performance of neural models is likely to be affected by the choice of hyper-parameters.
Approach:	They propose to automatically and dynamically determine batch sizes by accumulating gradients of mini-batches and performing an optimization step at just the time when the direction of gradients starts to fluctuate.
Outcome:	The proposed approach improves the Transformer model with a fixed 25k batch size by +0.73 and +0.82 BLEU respectively.

Copied to clipboard

Bo Xu, Erchen Yu, Jiahui Zhou, Hongfei Lin, Linlin Zong

Responsible & Trustworthy NLP

Challenge:	Existing models for multimodal hate detection lack implicit hateful cues, cross-modal-induced hate, and diversity of hate target groups.
Approach:	They propose a hypergraph-based prompting fusion model that uses LLMs to generate hate cue prompts and hypergraph learning to merge multimodal hate features.
Outcome:	The proposed model outperforms state-of-the-art models on two benchmark datasets showing that it can detect hate content across multiple modalities.

Paraphrasing as Zero-shot Translation with Feature-guided Diversity Enhancement (2026.acl-long)

Copied to clipboard

Ziyue Yan, Hongying Zan, Xinglin Lyu, Hongfei Xu

Text Generation

Challenge:	Existing studies use parallel corpora for training, which results in less diverse paraphrases.
Approach:	They train a bidirectional multilingual neural machine translation model on a bilingual parallel corpus and use it as a paraphrasing model.
Outcome:	The proposed method generates paraphrases with higher semantic consistency, literal fluency and sentential diversity than existing parabanks and LLMs.

Probing Word Translations in the Transformer and Trading Decoder for Encoder Layers (2021.naacl-main)

Copied to clipboard

Hongfei Xu, Josef van Genabith, Qiuhui Liu, Deyi Xiong

Breaking the Boundaries: A Unified Framework for Chinese Named Entity Recognition Across Text and Speech (2024.findings-emnlp)

Challenge:	Neural Machine Translation (NMT) has attracted wide attention in recent years.
Approach:	They propose a probing-based approach to measure word translation accuracy using transformer layers.
Outcome:	The proposed model outperforms previous probing-based translation models.

Copied to clipboard

Jinzhong Ning, Yuanyuan Sun, Bo Xu, Zhihao Yang, Ling Luo, Hongfei Lin

Multimodality

Challenge:	Existing approaches to Named Entity Recognition (NER) tasks are limited by the complexity of the data and the potential connections between tasks.
Approach:	They propose a task to break the boundaries between different modal NER tasks by using a unified data format for inputs from different modalités.
Outcome:	The proposed task breaks the boundaries between different modal NER tasks and is a unified implementation of them.

ParaZh-22M: A Large-Scale Chinese Parabank via Machine Translation (2022.coling-1)

Copied to clipboard

Wenjie Hao, Hongfei Xu, Deyi Xiong, Hongying Zan, Lingling Mu

Text Generation

Challenge:	Paraphrasing is an important data augmentation approach for natural language processing (NLP).
Approach:	They propose to extract sentence-level paraphrases from multiple Chinese translations and construct a larger Chinese parabank with 22M sentence pairs.
Outcome:	The proposed parabank is the largest to date in Chinese, but limited by one-to-many translation data.

Construction of a Chinese Corpus for the Analysis of the Emotionality of Metaphorical Expressions (P18-2)

Copied to clipboard

Dongyu Zhang, Hongfei Lin, Liang Yang, Shaowu Zhang, Bo Xu

Breakthrough from Nuance and Inconsistency: Enhancing Multimodal Sarcasm Detection with Context-Aware Self-Attention Fusion and Word Weight Calculation. (2024.lrec-main)

Challenge:	a corpus of 5,605 manually annotated sentences in Chinese is described . emotion is an abstract and vague conception, which is often described by metaphor .
Approach:	They propose to construct a corpus of metaphors annotated with emotion in Chinese . they use an annotation scheme to include linguistic metaphors, emotional categories and intensity .
Outcome:	The proposed corpus contains 5,605 manually annotated sentences in Chinese . the authors show that the corpus is large enough to analyze emotions .

Copied to clipboard

Hongfei Xue, Linyan Xu, Yu Tong, Rui Li, Jiali Lin, Dazhi Jiang

The Transference Architecture for Automatic Post-Editing (2020.coling-main)

Challenge:	Existing methods for sarcasm detection rely on feature concatenation to fuse different modalities or model inconsistencies among modalités.
Approach:	They propose to use Context-Aware Self-Attention Fusion to integrate local and momentary multimodal information into specific words to illustrate the inconsistencies between connotation and denotation.
Outcome:	The proposed method achieves an accuracy of 76.9 and an F1 score of 76.1 on the MUStARD dataset, surpassing the current state-of-the-art IWAN model by 1.7 and 1.6 respectively.

Copied to clipboard

Santanu Pal, Hongfei Xu, Nico Herbig, Sudip Kumar Naskar, Antonio Krüger, Josef van Genabith

Modeling Task-Aware MIMO Cardinality for Efficient Multilingual Neural Machine Translation (2021.acl-short)

Challenge:	A research challenge is the search for architectures that best support the capture, preparation and provision of src and mt information and its integration with pe decisions.
Approach:	They propose a multi-encoder based neural APE model that conditions post-editing decisions on both the source and machine translated text as inputs.
Outcome:	The proposed model outperforms the best performing systems by 1 BLEU point on the WMT 2016, 2017, and 2018 English–German APE shared tasks.

Copied to clipboard

Hongfei Xu, Qiuhui Liu, Josef van Genabith, Deyi Xiong

Multilinguality

Challenge:	Existing work has increased the modeling capacity of multilingual NMT by deepening or widening the Transformer.
Approach:	They propose to increase the model capacity by deepening the Transformer . they propose to use a multi-input-multi-output architecture to combine multiple inputs .
Outcome:	The proposed model surpasses previous work and is 1.31 times faster than existing models.

Cultural Bias Matters: A Cross-Cultural Benchmark Dataset and Sentiment-Enriched Model for Understanding Multimodal Metaphors (2025.acl-long)

Copied to clipboard

Senqi Yang, Dongyu Zhang, Jing Ren, Ziqi Xu, Xiuzhen Zhang, Yiliao Song, Hongfei Lin, Feng Xia

Rewiring the Transformer with Depth-Wise LSTMs (2024.lrec-main)

Challenge:	Metaphors are pervasive in communication, making them crucial for natural language processing.
Approach:	They propose a multicultural multimodal metaphor dataset designed for cross-cultural studies of metaphor in Chinese and English.
Outcome:	The proposed model improves metaphor comprehension across cultural backgrounds and cultural domains.

Copied to clipboard

Hongfei Xu, Yang Song, Qiuhui Liu, Josef van Genabith, Deyi Xiong

Self-Supervised Curriculum Learning for Spelling Error Correction (2021.emnlp-main)

Challenge:	Stacking non-linear layers allows deep neural networks to model complicated functions . but residual connections within each layer fail to fuse information from previous layers effectively .
Approach:	They propose a Transformer with depth-wise LSTMs connecting cascading Transformer layers and sub-layers.
Outcome:	The proposed model improves in English-German / French and multilingual tasks with BLEU.

Copied to clipboard

Zifa Gan, Hongfei Xu, Hongying Zan

Locality Preserving Sentence Encoding (2021.findings-emnlp)

Challenge:	Current approaches to SEC typically leverage a pre-training then fine-tuning procedure that treats data equally.
Approach:	They propose a self-supervised curriculum learning approach to improve model performance and model learning.
Outcome:	The proposed approach improves the model training and improves CL measurement.

Copied to clipboard

Changrong Min, Yonghe Chu, Liang Yang, Bo Xu, Hongfei Lin

Multi-Head Highly Parallelized LSTM Decoder for Neural Machine Translation (2021.acl-long)

Challenge:	Existing word embeddings can be used to learn sentence embedds on the sentence level.
Approach:	They propose a sentence embedding method that uses the inner product to compute semantic similarity between sentences.
Outcome:	The proposed method encodes sentences better in the sense of semantic structures.

Copied to clipboard

Hongfei Xu, Qiuhui Liu, Josef van Genabith, Deyi Xiong, Meng Zhang

WenetSpeech-Wu: Datasets, Benchmarks, and Models for a Unified Chinese Wu Dialect Speech Processing Ecosystem (2026.findings-acl)

Challenge:	a self-attention network can be easily parallelized at sequence level, but LSTMs are slower to train . a recent study shows that LS models require a lot of computations to perform .
Approach:	They propose to compute LSTMs at sequence level to enable sequence-level parallelization . they use a bag-of-words representation of the preceding tokens context to approximate LStms .
Outcome:	The proposed model performs better than existing models while being faster to train . the model can be trained efficiently due to the highly parallelized self-attention network .

Copied to clipboard

Chengyou Wang, Mingchen Shao, Jingbin Hu, Zeyu Zhu, Hongfei Xue, Bingshen Mu, Xin Xu, Xingyi Duan, Binbin Zhang, Zhu Pengcheng, Chuang Ding, Xiaojun Zhang, Hui Bu, Lei Xie

Multimodality

Challenge:	despite its linguistic significance, the Wu dialect of Chinese has long been hindered by the lack of large-scale speech data, standardized evaluation benchmarks, and publicly available models.
Approach:	They propose to use WenetSpeech-Wu as a large-scale, multi-dimensionally annotated open-source speech corpus for the Wu dialect of Chinese.
Outcome:	The proposed dataset includes 8,000 hours of speech data and strong open-source models . the proposed dataset is competitive and empirically validated .

Take Its Essence, Discard Its Dross! Debiasing for Toxic Language Detection via Counterfactual Causal Effect (2024.lrec-main)

Copied to clipboard

Junyu Lu, Bo Xu, Xiaokun Zhang, Kaiyuan Liu, Dongyu Zhang, Liang Yang, Hongfei Lin

Responsible & Trustworthy NLP

Challenge:	Existing methods to mitigate lexical bias in toxic language detection (TLD) do not exploit the “useful” and “misleading” impact of the bias.
Approach:	They propose a counterfactual Causal Debiasing Framework to mitigate lexical bias in toxic language detection (TLD) it preserves the “useful impact” of lexical bias and eliminates the "misleading impact" they propose to use the same framework to analyze the causal effect of a sentence and bias tokens .
Outcome:	The proposed framework preserves the “useful impact” of lexical bias and eliminates the ‘misleading impact’ Empirical evaluations show that the proposed model outperforms current debiased models for out-of-distribution data.

WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback (2026.acl-long)

Copied to clipboard

Taiwei Shi, Zhuoer Wang, Longqi Yang, Ying-Chun Lin, Zexue He, Mengting Wan, Pei Zhou, Sujay Kumar Jauhar, Sihao Chen, Shan Xia, Hongfei Zhang, Jieyu Zhao, Xiaofeng Xu, Xia Song, Jennifer Neville

Challenge:	Traditional alignment methods rely on human annotations and are subjective and misalignment with real-world user preferences.
Approach:	They propose a framework that leverages in-situ user feedback during conversations with LLMs to create preference datasets automatically.
Outcome:	The proposed framework identifies and classifies user feedback to LLM responses between conversation turns and creates examples of preferred and dispreferred responses according to user preferences.

Facilitating Fine-grained Detection of Chinese Toxic Language: Hierarchical Taxonomy, Resources, and Benchmarks (2023.acl-long)

Copied to clipboard

Junyu Lu, Bo Xu, Xiaokun Zhang, Changrong Min, Liang Yang, Hongfei Lin

Challenge:	Existing datasets suffer from a lack of fine-grained annotations, such as the toxic type and expressions with indirect toxicity.
Approach:	They propose a benchmark model to detect toxic language by incorporating lexical features into a Chinese dataset to facilitate fine-grained annotations.
Outcome:	The proposed model is based on insulting vocabulary containing implicit profanity and is able to detect toxic language with lexical features.

WECA: A WordNet-Encoded Collocation-Attention Network for Homographic Pun Recognition (D18-1)

Copied to clipboard

Yufeng Diao, Hongfei Lin, Di Wu, Liang Yang, Kan Xu, Zhihao Yang, Jian Wang, Shaowu Zhang, Bo Xu, Dongyu Zhang