Jiahao Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jiahao Xu has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Lihui Chen
Centre for Info. Sciences and Systems Nanyang Technological University
9 shared publications - Lemao Liu
Fudan University
4 shared publications - Wei Shao
City Univeristy of Hong Kong
4 shared publications - Charlie Soh
Nanyang Technological University Singapore
3 shared publications Cai Xinjun
Centre for Info. Sciences and Systems Nanyang Technological University
2 shared publicationsFangkun Wu
Centre for Info. Sciences and Systems Nanyang Technological University
2 shared publications
Latest Publications
- Open Access
FanChuan: A Multilingual and Graph-Structured Benchmark For Parody Detection and Analysis
Parody is an emerging phenomenon on social media, where individuals imitate a role or position opposite to their own, often for humor, provocation, or controversy. Detecting and analyzing parody can be challenging and is often reliant on context, yet it plays a crucial role in understanding cultural values, promoting subcultures, and enhancing self-expression. However, the study of parody is hindered by limited available data and deficient diversity in current datasets. To bridge this gap, we built seven parody datasets from both English and Chinese corpora, with 14,755 annotated users and 21,210 annotated comments in total. To provide sufficient context information, we also collect replies and construct user-interaction graphs to provide richer contextual information, which is lacking in existing datasets. With these datasets, we test traditional methods and Large Language Models (LLMs) on three key tasks: (1) parody detection, (2) comment sentiment analysis with parody, and (3) user sentiment analysis with parody. Our extensive experiments reveal that parody-related tasks still remain challenging for all models, and contextual information plays a critical role. Interestingly, we find that, in certain scenarios, traditional sentence embedding methods combined with simple classifiers can outperform advanced LLMs, i.e. DeepSeek-R1 and GPT-o3, highlighting parody as a significant challenge for LLMs.
Open AccessFanChuan: A Multilingual and Graph-Structured Benchmark For Parody Detection and Analysis
Parody is an emerging phenomenon on social media, where individuals imitate a role or position opposite to their own, often for humor, provocation, or controversy. Detecting and analyzing parody can be challenging and is often reliant on context, yet it plays a crucial role in understanding cultural values, promoting subcultures, and enhancing self-expression. However, the study of parody is hindered by limited available data and deficient diversity in current datasets. To bridge this gap, we built seven parody datasets from both English and Chinese corpora, with 14,755 annotated users and 21,210 annotated comments in total. To provide sufficient context information, we also collect replies and construct user-interaction graphs to provide richer contextual information, which is lacking in existing datasets. With these datasets, we test traditional methods and Large Language Models (LLMs) on three key tasks: (1) parody detection, (2) comment sentiment analysis with parody, and (3) user sentiment analysis with parody. Our extensive experiments reveal that parody-related tasks still remain challenging for all models, and contextual information plays a critical role. Interestingly, we find that, in certain scenarios, traditional sentence embedding methods combined with simple classifiers can outperform advanced LLMs, i.e. DeepSeek-R1 and GPT-o3, highlighting parody as a significant challenge for LLMs.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 21937-21957
FanChuan: A Multilingual and Graph-Structured Benchmark For Parody Detection and Analysis
Yilun Zheng, Sha Li, Fangkun Wu, Yang Ziyi, Lin Hongchao, Zhichao Hu, Cai Xinjun, Ziming Wang, Jinxuan Chen, Sitao Luan, Jiahao Xu, Lihui Chen. Findings of the Association for Computational Linguistics: ACL 2025. 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 21937-21957
Open AccessFanChuan: A Multilingual and Graph-Structured Benchmark For Parody Detection and Analysis
Yilun Zheng, Sha Li, Fangkun Wu, Yang Ziyi, Lin Hongchao, Zhichao Hu, Cai Xinjun, Ziming Wang, Jinxuan Chen, Sitao Luan, Jiahao Xu, Lihui Chen. Findings of the Association for Computational Linguistics: ACL 2025. 2025.
- Open Access
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 131-141
TransAgents: Build Your Translation Company with Language Agents
Minghao Wu, Jiahao Xu, Longyue Wang. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2024.
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 131-141
Open AccessTransAgents: Build Your Translation Company with Language Agents
Minghao Wu, Jiahao Xu, Longyue Wang. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2024.
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 43, 1-5
Research on Graph Neural Network Algorithm and Image Recognition Application
The emergence of graph neural networks makes up for the limitations of traditional neural networks in dealing with complex graph data, and they are widely used in the fields of social networks and recommender systems, but at the same time, there are some defects in themselves. This paper introduces the basics of graph neural networks, and also summarizes the advantages and disadvantages of GSN, GraphSAGE, DeepWalk and Node2Vec algorithms. Finally, it also discusses the application of graph neural networks in the field of image recognition, summarizes its deficiencies in practical applications, and puts forward the prospect of the future development of graph neural networks.
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 43, 1-5
Research on Graph Neural Network Algorithm and Image Recognition Application
The emergence of graph neural networks makes up for the limitations of traditional neural networks in dealing with complex graph data, and they are widely used in the fields of social networks and recommender systems, but at the same time, there are some defects in themselves. This paper introduces the basics of graph neural networks, and also summarizes the advantages and disadvantages of GSN, GraphSAGE, DeepWalk and Node2Vec algorithms. Finally, it also discusses the application of graph neural networks in the field of image recognition, summarizes its deficiencies in practical applications, and puts forward the prospect of the future development of graph neural networks.
- Open Access
DistillCSE: Distilled Contrastive Learning for Sentence Embeddings
This paper proposes the DistillCSE framework, which performs contrastive learning under the self-training paradigm with knowledge distillation. The potential advantage of DistillCSE is its self-enhancing feature: using a base model to provide additional supervision signals, a stronger model may be learned through knowledge distillation. However, the vanilla DistillCSE through the standard implementation of knowledge distillation only achieves marginal improvements due to severe overfitting. The further quantitative analyses demonstrate the reason that the standard knowledge distillation exhibits a relatively large variance of the teacher model's logits due to the essence of contrastive learning. To mitigate the issue induced by high variance, this paper accordingly proposed two simple yet effective solutions for knowledge distillation: a Group-P shuffling strategy as an implicit regularization and the averaging logits from multiple teacher components. Experiments on standard benchmarks demonstrate that the proposed DistillCSE outperforms many strong baseline methods and yields a new state-of-the-art performance.
Open AccessDistillCSE: Distilled Contrastive Learning for Sentence Embeddings
This paper proposes the DistillCSE framework, which performs contrastive learning under the self-training paradigm with knowledge distillation. The potential advantage of DistillCSE is its self-enhancing feature: using a base model to provide additional supervision signals, a stronger model may be learned through knowledge distillation. However, the vanilla DistillCSE through the standard implementation of knowledge distillation only achieves marginal improvements due to severe overfitting. The further quantitative analyses demonstrate the reason that the standard knowledge distillation exhibits a relatively large variance of the teacher model's logits due to the essence of contrastive learning. To mitigate the issue induced by high variance, this paper accordingly proposed two simple yet effective solutions for knowledge distillation: a Group-P shuffling strategy as an implicit regularization and the averaging logits from multiple teacher components. Experiments on standard benchmarks demonstrate that the proposed DistillCSE outperforms many strong baseline methods and yields a new state-of-the-art performance.
- Open Access
SimCSE++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives
This paper improves contrastive learning for sentence embeddings from two perspectives: handling dropout noise and addressing feature corruption. Specifically, for the first perspective, we identify that the dropout noise from negative pairs affects the model's performance. Therefore, we propose a simple yet effective method to deal with such type of noise. Secondly, we pinpoint the rank bottleneck of current solutions to feature corruption and propose a dimension-wise contrastive learning objective to address this issue. Both proposed methods are generic and can be applied to any contrastive learning based models for sentence embeddings. Experimental results on standard benchmarks demonstrate that combining both proposed methods leads to a gain of 1.8 points compared to the strong baseline SimCSE configured with BERT base. Furthermore, applying the proposed method to DiffCSE, another strong contrastive learning based baseline, results in a gain of 1.4 points.
Open AccessSimCSE++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives
This paper improves contrastive learning for sentence embeddings from two perspectives: handling dropout noise and addressing feature corruption. Specifically, for the first perspective, we identify that the dropout noise from negative pairs affects the model's performance. Therefore, we propose a simple yet effective method to deal with such type of noise. Secondly, we pinpoint the rank bottleneck of current solutions to feature corruption and propose a dimension-wise contrastive learning objective to address this issue. Both proposed methods are generic and can be applied to any contrastive learning based models for sentence embeddings. Experimental results on standard benchmarks demonstrate that combining both proposed methods leads to a gain of 1.8 points compared to the strong baseline SimCSE configured with BERT base. Furthermore, applying the proposed method to DiffCSE, another strong contrastive learning based baseline, results in a gain of 1.4 points.
- 11 October 2023
- Published by Elsevier in Expert Systems with Applications
- Vol. 238
BlendCSE: Blend contrastive learnings for sentence embeddings with rich semantics and transferability
- 11 October 2023
- Published by Elsevier in Expert Systems with Applications
- Vol. 238
BlendCSE: Blend contrastive learnings for sentence embeddings with rich semantics and transferability
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 8153-8165
DistillCSE: Distilled Contrastive Learning for Sentence Embeddings
Jiahao Xu, Wei Shao, Lihui Chen, Lemao Liu. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 8153-8165
Open AccessDistillCSE: Distilled Contrastive Learning for Sentence Embeddings
Jiahao Xu, Wei Shao, Lihui Chen, Lemao Liu. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 12028-12040
SimCSE++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives
Jiahao Xu, Wei Shao, Lihui Chen, Lemao Liu. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 12028-12040
Open AccessSimCSE++: Improving Contrastive Learning for Sentence Embeddings from Two Perspectives
Jiahao Xu, Wei Shao, Lihui Chen, Lemao Liu. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 2023.
- Open Access
- 1 December 2022
- Published by Wiley in Journal of Cachexia, Sarcopenia and Muscle
- Vol. 14 (1) , 356-368
MicroRNA profiling of different exercise interventions for alleviating skeletal muscle atrophy in naturally aging rats
Background: Exercise is an affordable and practical strategy to alleviate several detrimental outcomes from the aging process, including sarcopenia. The elucidation of molecular mechanisms to alleviate sarcopenia is one of the most important steps towards understanding human aging. Although microRNAs (miRNAs) regulate muscle growth, regeneration and aging, the potential role of exercise‐mediated miRNAs during the prevention and rehabilitation of skeletal muscle atrophy upon exercise interventions remains unclear.Methods: A miRNA profile by miRNA sequencing for gastrocnemius muscle of a 24‐month‐old aged male rat model mimicking the naturally aging process was established through screening the differentially expressed miRNAs (DEMs) for alleviating aging‐induced skeletal muscle atrophy upon optimal exercise intervention. The screened miRNAs and hub genes, as well as biomarkers with the most significantly enriched pathways, were validated by quantitative real‐time polymerase chain reaction and western blotting.Results: The sarcopenia index (SI) value and cross‐sectional area (CSA) of rats from the old control (OC) group significantly decreased when compared with the youth control (YC) group (P < 0.001, P < 0.01), whereas an increased SI value and an enlarged CSA of rats from the old‐aerobic exercise (OE), old‐resistance exercise (OR) and old‐mixed exercise (OM) groups were determined (P < 0.01, P < 0.001, P < 0.05; P < 0.01, P < 0.01, P < 0.05). Our results demonstrate that 764 known miRNAs, 201 novel miRNAs and 505 miRNA–mRNA interaction networks were identified to be related to aging‐induced muscular atrophy. Among them, 13 miRNAs were differentially expressed (P < 0.05 and log2|fold change| > 1) between the YC group and the OC group. Compared with the OC group, 7, 2 and 11 miRNAs were differentially expressed in the OE, OR and OM groups after exercise interventions, respectively. Meanwhile, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses revealed that the identified DEMs were primarily related to apoptosis, autophagy and the NF‐κB/MuRF1 signalling pathways (P < 0.05). Meanwhile, four DEMs (miR‐7a‐1‐3p, miR‐135a‐5p, miR‐151‐5p and miR‐196b‐5p), six hub genes (Ar, Igf1, Hif1a, Bdnf, Fak and Nras) and several biomarkers (LC3, Beclin1, p62, Bax, Bcl‐2 and NF‐κB/MuRF1) with the most significantly enriched pathways were confirmed, which may play a key role in muscular atrophy during the aging process.Conclusions: These findings are closely correlated with the progression of sarcopenia and could act as potential biomarkers for the diagnosis and interventional monitoring of aging‐induced skeletal muscle atrophy.
- 1 December 2022
- Published by Wiley in Journal of Cachexia, Sarcopenia and Muscle
- Vol. 14 (1) , 356-368
Open AccessMicroRNA profiling of different exercise interventions for alleviating skeletal muscle atrophy in naturally aging rats
Background: Exercise is an affordable and practical strategy to alleviate several detrimental outcomes from the aging process, including sarcopenia. The elucidation of molecular mechanisms to alleviate sarcopenia is one of the most important steps towards understanding human aging. Although microRNAs (miRNAs) regulate muscle growth, regeneration and aging, the potential role of exercise‐mediated miRNAs during the prevention and rehabilitation of skeletal muscle atrophy upon exercise interventions remains unclear.Methods: A miRNA profile by miRNA sequencing for gastrocnemius muscle of a 24‐month‐old aged male rat model mimicking the naturally aging process was established through screening the differentially expressed miRNAs (DEMs) for alleviating aging‐induced skeletal muscle atrophy upon optimal exercise intervention. The screened miRNAs and hub genes, as well as biomarkers with the most significantly enriched pathways, were validated by quantitative real‐time polymerase chain reaction and western blotting.Results: The sarcopenia index (SI) value and cross‐sectional area (CSA) of rats from the old control (OC) group significantly decreased when compared with the youth control (YC) group (P < 0.001, P < 0.01), whereas an increased SI value and an enlarged CSA of rats from the old‐aerobic exercise (OE), old‐resistance exercise (OR) and old‐mixed exercise (OM) groups were determined (P < 0.01, P < 0.001, P < 0.05; P < 0.01, P < 0.01, P < 0.05). Our results demonstrate that 764 known miRNAs, 201 novel miRNAs and 505 miRNA–mRNA interaction networks were identified to be related to aging‐induced muscular atrophy. Among them, 13 miRNAs were differentially expressed (P < 0.05 and log2|fold change| > 1) between the YC group and the OC group. Compared with the OC group, 7, 2 and 11 miRNAs were differentially expressed in the OE, OR and OM groups after exercise interventions, respectively. Meanwhile, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses revealed that the identified DEMs were primarily related to apoptosis, autophagy and the NF‐κB/MuRF1 signalling pathways (P < 0.05). Meanwhile, four DEMs (miR‐7a‐1‐3p, miR‐135a‐5p, miR‐151‐5p and miR‐196b‐5p), six hub genes (Ar, Igf1, Hif1a, Bdnf, Fak and Nras) and several biomarkers (LC3, Beclin1, p62, Bax, Bcl‐2 and NF‐κB/MuRF1) with the most significantly enriched pathways were confirmed, which may play a key role in muscular atrophy during the aging process.Conclusions: These findings are closely correlated with the progression of sarcopenia and could act as potential biomarkers for the diagnosis and interventional monitoring of aging‐induced skeletal muscle atrophy.