🧑‍🔬 Xiangyu Qi

🏛️ Affiliations

🏛️Jiangsu Academy of Agricultural SciencesNanjing · China2020–2026 · 25 paperscurrent🏥Shandong Provincial HospitalJinan · China2020–2026 · 10 paperscurrent🎓Shandong First Medical UniversityJinan · China2022–2026 · 6 paperscurrent🎓Northwest A&F UniversityYangling · China2022–2026 · 3 papersEmployment since 2022 · 学生, 草业与草原学院current🎓Princeton UniversityPrinceton · United States2022–2025 · 15 papers🎓Nanjing Agricultural UniversityNanjing · China2013–2022 · 14 papers🎓Shandong UniversityJinan · China2020–2023 · 7 papers🏛️Ministry of Agriculture and Rural AffairsBeijing · China2021–2022 · 3 papers🎓Beijing University of AgricultureBeijing · China2022 · 2 papers

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 3 more institutions on single bylines are not listed.

🧭 Fields, topics and concepts

Of its works in our library, with the number of works on each.

📅 Works by year

2020s

📄 Papers

The most cited first (OpenAlex and our library).

1. Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To! (2023 · 355 citations)
   Current safety alignment methods, such as instruction tuning and RLHF, constrain how a model behaves at inference time.
2. Visual Adversarial Examples Jailbreak Aligned Large Language Models (2023 · 240 citations)
   Adding image inputs to large language models creates new security risks. The authors argue that images are continuous and high-dimensional, which makes them much easier to manipulate adversarially than discrete text. Because these models can do so many things, a successful attack can do more than cause a…
3. Safety Alignment Should Be Made More Than Just a Few Tokens Deep (2024 · 97 citations)
   The paper asks why safety-aligned LLMs are so easy to break, whether through adversarial suffixes, prefilling, tweaked decoding parameters, or a few steps of fine-tuning.
4. SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal (2024 · 87 citations)
   SORRY-Bench is a benchmark that measures how reliably aligned LLMs decline unsafe requests. The authors identify three weaknesses in earlier safety evaluations. Their category schemes are coarse and inconsistent, and they concentrate on a few topics such as fraud while barely testing others like self-harm. They…
5. Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications (2024 · 66 citations)
   Aligned LLMs can be jailbroken or lose their safeguards even after harmless fine-tuning. The authors ask whether this fragility can be traced to specific parts of the weights. Using the Llama2-chat family, they attribute behaviour to individual neurons and to low-rank components of weight matrices. They use two…

All works

7 works of theirs in our library, the most cited first; 5 with a page here, the others link their original until we summarise them.

1. Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!arXiv:2310.03693Xiangyu Qi, Yi Zeng, Tinghao Xie, et al.2023 · 355 citations
2. Visual Adversarial Examples Jailbreak Aligned Large Language ModelsProceedings of the AAAI Conference on Artificial IntelligenceXiangyu Qi, Kaixuan Huang, Ashwinee Panda, et al.2023 · 240 citations
3. Safety Alignment Should Be Made More Than Just a Few Tokens DeeparXiv:2406.05946Xiangyu Qi, Ashwinee Panda, Kaifeng Lyu, et al.2024 · 97 citations
4. SORRY-Bench: Systematically Evaluating Large Language Model Safety RefusalarXiv:2406.14598Tinghao Xie, Xiangyu Qi, Yi Zeng, et al.2024 · 87 citations
5. Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank ModificationsarXiv:2402.05162Boyi Wei, Kaixuan Huang, Yangsibo Huang, et al.2024 · 66 citations
6. AI Risk Management Should Incorporate Both Safety and Security ↗arXiv (Cornell University) · no summary yetXiangyu Qi et al.2024 · 3 citations
7. The Model Hears You: Audio Language Model Deployments Should Consider the Principle of Least Privilege ↗Proceedings of the AAAI/ACM Conference on AI Ethics and Society · no summary yetLuxi He et al.2025 · 1 citation

🤝 Co-authors here

12 of 17 co-authors with a page, the most papers together first.

• Peter Henderson ×4
• Prateek Mittal ×4
• Kaixuan Huang ×3
• Tinghao Xie ×3
• Ashwinee Panda ×2
• Boyi Wei ×2
• Mengdi Wang ×2
• Ruoxi Jia ×2
• Yangsibo Huang ×2
• Yi Zeng ×2
• Bo Li
• Dacheng Li

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

• Proceedings of the AAAI Conference on Artificial Intelligence: Proceedings of the AAAI Conference on Artificial Intelligence 2023
• Proceedings of the AAAI/ACM Conference on AI Ethics and Society

🔗 Cited by

Papers of our library cite their works 905 times (of the 149,439 papers whose bibliographies we read); 215 citing papers with a page here →.

Updated 2 November 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/xiangyu-qi/