Xiang Yue
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
Xiang Yue has a total of 129 co-authors. Here are the top co-authors by number of shared publications:
- Wen Zhang
College of Informatics, Huazhong Agricultural University, Wuhan, China
9 shared publications - Huan Sun
The Ohio State University, Columbus, OH, USA
5 shared publications - Minxin Du
Department of Information Engineering, The Chinese University of Hong Kong, Hong Kong
3 shared publications Simon Lin
Abigail Wexner Research Institute at Nationwide Children’s Hospital
3 shared publications- Yanlin Chen
School of Mathematics and Statistics, Wuhan University, Wuhan 430072, China
3 shared publications Feng Huang
School of Computer Science, Wuhan University, Wuhan, China
2 shared publications
Latest Publications
- Open Access
- 29 September 2026
- Published by Springer Nature in Research Square
On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists
As AI capabilities advance, AI reviewers (agents that read a manuscript and produce a referee report) are beginning to be deployed in scientific peer review. Yet, their credibility remains in question: many scientists think that they cannot judge scientific validity, while others are more optimistic about their readiness. In this work, we conduct a study in which 45 scientists in Physical, Biological, and Health Sciences spend 469 hours rating 2,960 individual criticisms (each targeting one aspect of a paper) from human-written and AI-generated reviews of 82 Nature-family papers, scoring each on correctness, significance, and evidence sufficiency. On a composite rating of reviews across all three dimensions, an AI reviewer powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs 48.2% of criticisms, p=0.009). Also, AI reviewers' correct criticisms are more often rated significant, and surface a distinct 26% of legitimate issues that no human raises. However, AI reviewers are more often incorrect, overlap with each other far more than humans do with each other, and exhibit 16 weaknesses rare in human reviews. Overall, our results position current AI reviewers as complements to, not substitutes for, human reviewers.
- 29 September 2026
- Published by Springer Nature in Research Square
Open AccessOn the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists
As AI capabilities advance, AI reviewers (agents that read a manuscript and produce a referee report) are beginning to be deployed in scientific peer review. Yet, their credibility remains in question: many scientists think that they cannot judge scientific validity, while others are more optimistic about their readiness. In this work, we conduct a study in which 45 scientists in Physical, Biological, and Health Sciences spend 469 hours rating 2,960 individual criticisms (each targeting one aspect of a paper) from human-written and AI-generated reviews of 82 Nature-family papers, scoring each on correctness, significance, and evidence sufficiency. On a composite rating of reviews across all three dimensions, an AI reviewer powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs 48.2% of criticisms, p=0.009). Also, AI reviewers' correct criticisms are more often rated significant, and surface a distinct 26% of legitimate issues that no human raises. However, AI reviewers are more often incorrect, overlap with each other far more than humans do with each other, and exhibit 16 weaknesses rare in human reviews. Overall, our results position current AI reviewers as complements to, not substitutes for, human reviewers.
- 1 May 2026
- Published by Springer Nature
- p. 167-190
Unlearning Faithfulness: Unlearning Isn’t Deletion
- Open Access
- 15 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 2665-2679
DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Pass
Differentially private stochastic gradient descent (DP-SGD) adds noise to gradients in back-propagation, safeguarding training data from privacy leakage, particularly membership inference. It fails to cover (inference-time) threats like embedding inversion and sensitive attribute inference. It is also costly in storage and computation when used to fine-tune large pre-trained language models (LMs). We propose DP-Forward, which directly perturbs embedding matrices in the forward pass of LMs. It satisfies stringent local DP requirements for training and inference data. To instantiate it using the smallest matrix-valued noise, we devise an analytic matrix Gaussian mechanism (aMGM) by drawing possibly non-i.i.d. noise from a matrix Gaussian distribution. We then investigate perturbing outputs from different hidden (sub-)layers of LMs with aMGM noises. Its utility on three typical tasks almost hits the non-private baseline and outperforms DP-SGD by up to 7.7pp at a moderate privacy level. It saves 3x time and memory costs compared to DP-SGD with the latest high-speed library. It also reduces the average success rates of embedding inversion and sensitive attribute inference by up to 88pp and 41pp, respectively, whereas DP-SGD fails.
- 15 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 2665-2679
Open AccessDP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Pass
Differentially private stochastic gradient descent (DP-SGD) adds noise to gradients in back-propagation, safeguarding training data from privacy leakage, particularly membership inference. It fails to cover (inference-time) threats like embedding inversion and sensitive attribute inference. It is also costly in storage and computation when used to fine-tune large pre-trained language models (LMs). We propose DP-Forward, which directly perturbs embedding matrices in the forward pass of LMs. It satisfies stringent local DP requirements for training and inference data. To instantiate it using the smallest matrix-valued noise, we devise an analytic matrix Gaussian mechanism (aMGM) by drawing possibly non-i.i.d. noise from a matrix Gaussian distribution. We then investigate perturbing outputs from different hidden (sub-)layers of LMs with aMGM noises. Its utility on three typical tasks almost hits the non-private baseline and outperforms DP-SGD by up to 7.7pp at a moderate privacy level. It saves 3x time and memory costs compared to DP-SGD with the latest high-speed library. It also reduces the average success rates of embedding inversion and sensitive attribute inference by up to 88pp and 41pp, respectively, whereas DP-SGD fails.
- Open Access
MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning
We introduce MAmmoTH, a series of open-source large language models (LLMs) specifically tailored for general math problem-solving. The MAmmoTH models are trained on MathInstruct, our meticulously curated instruction tuning dataset. MathInstruct is compiled from 13 math datasets with intermediate rationales, six of which have rationales newly curated by us. It presents a unique hybrid of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and also ensures extensive coverage of diverse fields in math. The hybrid of CoT and PoT not only unleashes the potential of tool use but also allows different thought processes for different math problems. As a result, the MAmmoTH series substantially outperform existing open-source models on nine mathematical reasoning datasets across all scales with an average accuracy gain between 16% and 32%. Remarkably, our MAmmoTH-7B model reaches 33% on MATH (a competition-level dataset), which exceeds the best open-source 7B model (WizardMath) by 23%, and the MAmmoTH-34B model achieves 44% accuracy on MATH, even surpassing GPT-4's CoT result. Our work underscores the importance of diverse problem coverage and the use of hybrid rationales in developing superior math generalist models.
Open AccessMAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuning
We introduce MAmmoTH, a series of open-source large language models (LLMs) specifically tailored for general math problem-solving. The MAmmoTH models are trained on MathInstruct, our meticulously curated instruction tuning dataset. MathInstruct is compiled from 13 math datasets with intermediate rationales, six of which have rationales newly curated by us. It presents a unique hybrid of chain-of-thought (CoT) and program-of-thought (PoT) rationales, and also ensures extensive coverage of diverse fields in math. The hybrid of CoT and PoT not only unleashes the potential of tool use but also allows different thought processes for different math problems. As a result, the MAmmoTH series substantially outperform existing open-source models on nine mathematical reasoning datasets across all scales with an average accuracy gain between 16% and 32%. Remarkably, our MAmmoTH-7B model reaches 33% on MATH (a competition-level dataset), which exceeds the best open-source 7B model (WizardMath) by 23%, and the MAmmoTH-34B model achieves 44% accuracy on MATH, even surpassing GPT-4's CoT result. Our work underscores the importance of diverse problem coverage and the use of hybrid rationales in developing superior math generalist models.
- Open Access
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 2349-2359
Sanitizing Sentence Embeddings (and Labels) for Local Differential Privacy
Differentially private (DP) learning, notably DP stochastic gradient descent (DP-SGD), has limited applicability in fine-tuning gigantic pre-trained language models (LMs) for natural language processing tasks. The culprit is the perturbation of gradients (as gigantic as entire models), leading to significant efficiency and accuracy drops. We show how to achieve metric-based local DP (LDP) by sanitizing (high-dimensional) sentence embedding, extracted by LMs and much smaller than gradients. For potential utility improvement, we impose a consistency constraint on the sanitization. We explore two approaches: One is brand new and can directly output consistent noisy embeddings; the other is an upgradation with post-processing. To further mitigate “the curse of dimensionality,” we introduce two trainable linear maps for mediating dimensions without hurting privacy or utility. Our protection can effectively defend against privacy threats on embeddings. It also naturally extends to inference. Our experiments1 show that we reach the non-private accuracy under properly configured parameters, e.g., 0.92 for SST-2 with a privacy budget ϵ = 10 and the reduced dimension as 16. We also sanitize the label for LDP (with another small privacy budget) with limited accuracy losses to fully protect every sequence-label pair.
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 2349-2359
Open AccessSanitizing Sentence Embeddings (and Labels) for Local Differential Privacy
Differentially private (DP) learning, notably DP stochastic gradient descent (DP-SGD), has limited applicability in fine-tuning gigantic pre-trained language models (LMs) for natural language processing tasks. The culprit is the perturbation of gradients (as gigantic as entire models), leading to significant efficiency and accuracy drops. We show how to achieve metric-based local DP (LDP) by sanitizing (high-dimensional) sentence embedding, extracted by LMs and much smaller than gradients. For potential utility improvement, we impose a consistency constraint on the sanitization. We explore two approaches: One is brand new and can directly output consistent noisy embeddings; the other is an upgradation with post-processing. To further mitigate “the curse of dimensionality,” we introduce two trainable linear maps for mediating dimensions without hurting privacy or utility. Our protection can effectively defend against privacy threats on embeddings. It also naturally extends to inference. Our experiments1 show that we reach the non-private accuracy under properly configured parameters, e.g., 0.92 for SST-2 with a privacy budget ϵ = 10 and the reduced dimension as 16. We also sanitize the label for LDP (with another small privacy budget) with limited accuracy losses to fully protect every sequence-label pair.
- Open Access
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 197-201
Roll Up Your Sleeves: Working with a Collaborative and Engaging Task-Oriented Dialogue System
Lingbo Mo, Shijie Chen, Ziru Chen, Xiang Deng, Ashley Lewis, Sunit Singh, Samuel Stevens, Chang-You Tai, Zhen Wang, Xiang Yue, Tianshu Zhang, Yu Su, Huan Sun. Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue. 2023.
- 1 January 2023
- Published by Association for Computational Linguistics (ACL)
- p. 197-201
Open AccessRoll Up Your Sleeves: Working with a Collaborative and Engaging Task-Oriented Dialogue System
Lingbo Mo, Shijie Chen, Ziru Chen, Xiang Deng, Ashley Lewis, Sunit Singh, Samuel Stevens, Chang-You Tai, Zhen Wang, Xiang Yue, Tianshu Zhang, Yu Su, Huan Sun. Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue. 2023.
- Open Access
C-MORE: Pretraining to Answer Open-Domain Questions by Consulting Millions of References
We consider the problem of pretraining a two-stage open-domain question answering (QA) system (retriever + reader) with strong transfer capabilities. The key challenge is how to construct a large amount of high-quality question-answer-context triplets without task-specific annotations. Specifically, the triplets should align well with downstream tasks by: (i) covering a wide range of domains (for open-domain applications), (ii) linking a question to its semantically relevant context with supporting evidence (for training the retriever), and (iii) identifying the correct answer in the context (for training the reader). Previous pretraining approaches generally fall short of one or more of these requirements. In this work, we automatically construct a large-scale corpus that meets all three criteria by consulting millions of references cited within Wikipedia. The well-aligned pretraining signals benefit both the retriever and the reader significantly. Our pretrained retriever leads to 2%-10% absolute gains in top-20 accuracy. And with our pretrained reader, the entire system improves by up to 4% in exact match.
Open AccessC-MORE: Pretraining to Answer Open-Domain Questions by Consulting Millions of References
We consider the problem of pretraining a two-stage open-domain question answering (QA) system (retriever + reader) with strong transfer capabilities. The key challenge is how to construct a large amount of high-quality question-answer-context triplets without task-specific annotations. Specifically, the triplets should align well with downstream tasks by: (i) covering a wide range of domains (for open-domain applications), (ii) linking a question to its semantically relevant context with supporting evidence (for training the retriever), and (iii) identifying the correct answer in the context (for training the reader). Previous pretraining approaches generally fall short of one or more of these requirements. In this work, we automatically construct a large-scale corpus that meets all three criteria by consulting millions of references cited within Wikipedia. The well-aligned pretraining signals benefit both the retriever and the reader significantly. Our pretrained retriever leads to 2%-10% absolute gains in top-20 accuracy. And with our pretrained reader, the entire system improves by up to 4% in exact match.
- Open Access
- 9 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 580-587
CliniQG4QA: Generating Diverse Questions for Domain Adaptation of Clinical Question Answering
Clinical question answering (QA) aims to automatically answer questions from medical professionals based on clinical texts. Studies show that neural QA models trained on one corpus may not generalize well to new clinical texts from a different institute or a different patient group, where largescale QA pairs are not readily available for model retraining. To address this challenge, we propose a simple yet effective framework, CliniQG4QA, which leverages question generation (QG) to synthesize QA pairs on new clinical contexts and boosts QA models without requiring manual annotations. In order to generate diverse types of questions that are essential for training QA models, we further introduce a seq2seq-based question phrase prediction (QPP) module that can be used together with most existing QG models to diversify the generation. Our comprehensive experiment results show that the QA corpus generated by our framework can improve QA models on the new contexts (up to 8% absolute gain in terms of Exact Match), and that the QPP module plays a crucial role in achieving the gain. 1 1 Our dataset and code are available at: https://github.com/sunlabosu/CliniQG4QA/.
- 9 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 580-587
Open AccessCliniQG4QA: Generating Diverse Questions for Domain Adaptation of Clinical Question Answering
Clinical question answering (QA) aims to automatically answer questions from medical professionals based on clinical texts. Studies show that neural QA models trained on one corpus may not generalize well to new clinical texts from a different institute or a different patient group, where largescale QA pairs are not readily available for model retraining. To address this challenge, we propose a simple yet effective framework, CliniQG4QA, which leverages question generation (QG) to synthesize QA pairs on new clinical contexts and boosts QA models without requiring manual annotations. In order to generate diverse types of questions that are essential for training QA models, we further introduce a seq2seq-based question phrase prediction (QPP) module that can be used together with most existing QG models to diversify the generation. Our comprehensive experiment results show that the QA corpus generated by our framework can improve QA models on the new contexts (up to 8% absolute gain in terms of Exact Match), and that the QPP module plays a crucial role in achieving the gain. 1 1 Our dataset and code are available at: https://github.com/sunlabosu/CliniQG4QA/.
- Open Access
- 1 January 2021
- Published by Association for Computational Linguistics (ACL)
- p. 3759-3769
COUGH: A Challenge Dataset and Models for COVID-19 FAQ Retrieval
Xinliang Frederick Zhang, Heming Sun, Xiang Yue, Simon Lin, Huan Sun. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021.
- 1 January 2021
- Published by Association for Computational Linguistics (ACL)
- p. 3759-3769
Open AccessCOUGH: A Challenge Dataset and Models for COVID-19 FAQ Retrieval
Xinliang Frederick Zhang, Heming Sun, Xiang Yue, Simon Lin, Huan Sun. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021.
- Open Access
PHICON: Improving Generalization of Clinical Text De-identification Models via Data Augmentation
De-identification is the task of identifying protected health information (PHI) in the clinical text. Existing neural de-identification models often fail to generalize to a new dataset. We propose a simple yet effective data augmentation method PHICON to alleviate the generalization issue. PHICON consists of PHI augmentation and Context augmentation, which creates augmented training corpora by replacing PHI entities with named-entities sampled from external sources, and by changing background context with synonym replacement or random word insertion, respectively. Experimental results on the i2b2 2006 and 2014 de-identification challenge datasets show that PHICON can help three selected de-identification models boost F1-score (by at most 8.6%) on cross-dataset test setting. We also discuss how much augmentation to use and how each augmentation method influences the performance.
Open AccessPHICON: Improving Generalization of Clinical Text De-identification Models via Data Augmentation
De-identification is the task of identifying protected health information (PHI) in the clinical text. Existing neural de-identification models often fail to generalize to a new dataset. We propose a simple yet effective data augmentation method PHICON to alleviate the generalization issue. PHICON consists of PHI augmentation and Context augmentation, which creates augmented training corpora by replacing PHI entities with named-entities sampled from external sources, and by changing background context with synonym replacement or random word insertion, respectively. Experimental results on the i2b2 2006 and 2014 de-identification challenge datasets show that PHICON can help three selected de-identification models boost F1-score (by at most 8.6%) on cross-dataset test setting. We also discuss how much augmentation to use and how each augmentation method influences the performance.