Hanqing Guo

Dr. Hanqing Guo


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
14 April 2025 in Remote Sensing
Verified
Journal Article
Open access
PDF file
YUFEI CHUYUFEI CHU
YUFEI CHU
GLGuo Lin
Guo Lin
MDMin Deng
Min Deng
Hanqing GuoHanqing Guo
Hanqing Guo
Jun ZhangJun Zhang
Jun Zhang
As machine learning becomes more integrated into atmospheric science, XGBoost has gained popularity for its ability to assess the relative contributions of influencing factors in the atmospheric boundary layer height. To examine how these factors vary across seasons, a seasonal analysis is necessary. However, dividing data by season reduces the sample size, which can affect result reliability and complicate factor comparisons. To address these challenges, this study replaces default parameters with grid search optimization and incorporates cross-validation to mitigate dataset limitations. Using XGBoost with four years of data from the atmospheric radiation measurement (ARM) (Southern Great Plains (SGP) C1 site, cross-validation stabilizes correlation coefficient fluctuations from 0.3 to within 0.1. With optimized parameters, the R value can reach 0.81. Analysis of the C1 site reveals that the relative importance of different factors changes across seasons. Lower tropospheric stability (LTS, ~0.53) is the dominant factor at C1 throughout the year. However, during DJF, latent heat flux (LHF, 0.44) surpasses LTS (0.22). In SON, LTS (0.58) becomes more influential than LHF (0.18). Further comparisons among the four long-term SGP sites (C1, E32, E37, and E39) show seasonal variations in relative importance. Notably, during JJA, the differences in the relative importance of the three factors across all sites are lower than in other seasons. This suggests that boundary layer development in the summer is not dominated by a single factor, reflecting a more intricate process likely influenced by seasonal conditions such as enhanced convective activity, higher temperatures, and humidity, which collectively contribute to a balanced distribution of parameter impacts. Furthermore, the relative importance of LTS gradually increases from morning to noon, indicating that LTS becomes more significant as the boundary layer approaches its maximum height. Consequently, the LTS in the early morning in autumn exhibits greater relative importance compared to other seasons. This reflects a faster development of the mixing layer height (MLH) in autumn, suggesting that it is easier to retrieve the MLH from the previous day during this period. The findings enhance understanding of boundary layer evolution and contribute to improved boundary layer parameterization.

1 January 2025
Verified
Conference Paper
Open access
PDF file
YZYisheng Zhong
Yisheng Zhong
YWYizhu Wen
Yizhu Wen
JGJunfeng Guo
Junfeng Guo
MKMehran Kafai
Mehran Kafai
HHHeng Huang
Heng Huang
Zhuangdi ZhuZhuangdi Zhu
Zhuangdi Zhu
Yisheng Zhong, Yizhu Wen, Junfeng Guo, Mehran Kafai, Heng Huang, Hanqing Guo, Zhuangdi Zhu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

10 December 2024 in arXiv
Verified
Preprint
Open access
PDF file
BCBocheng Chen
Bocheng Chen
Hanqing GuoHanqing Guo
Hanqing Guo
QYQiben Yan
Qiben Yan
Defense in large language models (LLMs) is crucial to counter the numerous attackers exploiting these systems to generate harmful content through manipulated prompts, known as jailbreak attacks. Although many defense strategies have been proposed, they often require access to the model's internal structure or need additional training, which is impractical for service providers using LLM APIs, such as OpenAI APIs or Claude APIs. In this paper, we propose a moving target defense approach that alters decoding hyperparameters to enhance model robustness against various jailbreak attacks. Our approach does not require access to the model's internal structure and incurs no additional training costs. The proposed defense includes two key components: (1) optimizing the decoding strategy by identifying and adjusting decoding hyperparameters that influence token generation probabilities, and (2) transforming the decoding hyperparameters and model system prompts into dynamic targets, which are continuously altered during each runtime. By continuously modifying decoding strategies and prompts, the defense effectively mitigates the existing attacks. Our results demonstrate that our defense is the most effective against jailbreak attacks in three of the models tested when using LLMs as black-box APIs. Moreover, our defense offers lower inference costs and maintains comparable response quality, making it a potential layer of protection when used alongside other defense methods.

24 May 2025 in arXiv
Verified
Preprint
Open access
PDF file
BMBinhao Ma
Binhao Ma
Hanqing GuoHanqing Guo
Hanqing Guo
ZLZhengping Jay Luo
Zhengping Jay Luo
RDRui Duan
Rui Duan
Recent advances in Multimodal Large Language Models (MLLMs) have significantly enhanced the naturalness and flexibility of human computer interaction by enabling seamless understanding across text, vision, and audio modalities. Among these, voice enabled models such as SpeechGPT have demonstrated considerable improvements in usability, offering expressive, and emotionally responsive interactions that foster deeper connections in real world communication scenarios. However, the use of voice introduces new security risks, as attackers can exploit the unique characteristics of spoken language, such as timing, pronunciation variability, and speech to text translation, to craft inputs that bypass defenses in ways not seen in text-based systems. Despite substantial research on text based jailbreaks, the voice modality remains largely underexplored in terms of both attack strategies and defense mechanisms. In this work, we present an adversarial attack targeting the speech input of aligned MLLMs in a white box scenario. Specifically, we introduce a novel token level attack that leverages access to the model's speech tokenization to generate adversarial token sequences. These sequences are then synthesized into audio prompts, which effectively bypass alignment safeguards and to induce prohibited outputs. Evaluated on SpeechGPT, our approach achieves up to 89 percent attack success rate across multiple restricted tasks, significantly outperforming existing voice based jailbreak methods. Our findings shed light on the vulnerabilities of voice-enabled multimodal systems and to help guide the development of more robust next-generation MLLMs.

1 September 2024 in arXiv
Verified
Preprint
Open access
PDF file
BCBocheng Chen
Bocheng Chen
Hanqing GuoHanqing Guo
Hanqing Guo
GWGuangjing Wang
Guangjing Wang
YWYuanda Wang
Yuanda Wang
QYQiben Yan
Qiben Yan
Large Language Models (LLMs) have demonstrated great capabilities in natural language understanding and generation, largely attributed to the intricate alignment process using human feedback. While alignment has become an essential training component that leverages data collected from user queries, it inadvertently opens up an avenue for a new type of user-guided poisoning attacks. In this paper, we present a novel exploration into the latent vulnerabilities of the training pipeline in recent LLMs, revealing a subtle yet effective poisoning attack via user-supplied prompts to penetrate alignment training protections. Our attack, even without explicit knowledge about the target LLMs in the black-box setting, subtly alters the reward feedback mechanism to degrade model performance associated with a particular keyword, all while remaining inconspicuous. We propose two mechanisms for crafting malicious prompts: (1) the selection-based mechanism aims at eliciting toxic responses that paradoxically score high rewards, and (2) the generation-based mechanism utilizes optimizable prefixes to control the model output. By injecting 1\% of these specially crafted prompts into the data, through malicious users, we demonstrate a toxicity score up to two times higher when a specific trigger word is used. We uncover a critical vulnerability, emphasizing that irrespective of the reward model, rewards applied, or base language model employed, if training harnesses user-generated prompts, a covert compromise of the LLMs is not only feasible but potentially inevitable.