User:Linyueqian
 | This user has publicly declared that they have a conflict of interest regarding the Wikipedia article Yiran Chen. |
| This is a Wikipedia user page. This is not an encyclopedia article or the talk page for an encyclopedia article. If you find this page on any site other than Wikipedia, you are viewing a mirror site. Be aware that the page may be outdated and that the user whom this page is about may have no personal affiliation with any site other than Wikipedia. The original page is located at https://en.wikipedia.org/wiki/User:Linyueqian. |
Yueqian Lin, also known as Max Lin, is a computer scientist and doctoral candidate at Duke University, known for his research in multimodal large language models, audio and vision understanding, and singing voice synthesis. He is a committer of the open-source project vLLM-Omni.[1]
Education
[edit]Lin earned a Bachelor of Science degree in Data Science jointly from Duke Kunshan University (DKU) and Duke University in May 2024, graduating summa cum laude.[2] He received a signature work distinction and served as the valedictorian of the Class of 2024 at Duke Kunshan University.[3] His commencement speech drew on themes from the Tao Te Ching, reflecting on the significance of being the third graduating class of the university.[4]
In August 2024, Lin began his Ph.D. studies in the Department of Electrical and Computer Engineering at Duke University, where he is advised by Yiran Chen and Hai "Helen" Li in the Center for Computational Evolutionary Intelligence (CEI).[2] He passed his qualifying exam in August 2025 and his preliminary exam in February 2026, becoming a Ph.D. candidate.[5]
Research
[edit]Lin's research focuses on multimodal large language models, with emphasis on audio and vision understanding and generation. His work spans areas including token pruning for efficient inference, singing voice synthesis, adversarial robustness, and out-of-distribution detection.
Audio and speech processing
[edit]Lin co-developed SpeechPrune, a context-aware token pruning method for speech information retrieval, along with the SPIRAL benchmark dataset for evaluating long-form spoken content comprehension. The work was presented as an oral presentation at IEEE ICME 2025.[5]
His earlier work on singing voice synthesis includes BiSinger, a bilingual singing voice synthesis system capable of handling Chinese, English, and code-switching, published at IEEE ASRU 2023.[6] He also co-introduced the ACE-Opencpop and KiSing-v2 large-scale singing voice datasets, presented as an oral paper at Interspeech 2024.[7]
Computer vision and multimodal systems
[edit]Lin co-authored HippoMM, a hippocampal-inspired multimodal memory system for long audiovisual event understanding, accepted at the CVPR 2026 Findings Track.[8]
His work on SD-NAE (Stable Diffusion Natural Adversarial Examples) proposed a method for generating natural adversarial examples using Stable Diffusion, presented at ICLR 2024.[9]
He also contributed to Vision-Zero, a scalable method for vision-language model self-improvement through gamified self-play, accepted at ICLR 2026,[5] and CoreMatching, a co-adaptive sparse inference framework accepted at ICML 2025.[5]
AI safety
[edit]Lin co-developed MUSE, an open-source platform for multimodal unified safety evaluation and red-teaming of large language models, which integrates automated cross-modal attack generation and a multi-turn adversarial evaluation framework.[10]
Industry experience
[edit]In the summer of 2025, Lin was a research intern at Adobe Research in San Jose, California.[5]
Professional memberships
[edit]Lin is a member of the IEEE, the ACM, and Sigma Xi.[5]
Selected publications
[edit]- Wang, Z., Lin, Y., et al. "MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Large Language Models." arXiv preprint arXiv:2603.02482 (2026).
- Lin, Y., et al. "HippoMM: Hippocampal-inspired Multimodal Memory for Long Audiovisual Event Understanding." CVPR Findings (2026).
- Lin, Y., et al. "VERA: Voice Evaluation of Reasoning Ability." ICLR Workshop (2026).
- Lin, Y., et al. "SpeechPrune: Context-aware Token Pruning for Speech Information Retrieval." IEEE ICME (2025).
- Shi, J., Lin, Y., et al. "Singing Voice Data Scaling-up: An Introduction to ACE-Opencpop and KiSing-v2." Interspeech (2024).
- Lin, Y., et al. "SD-NAE: Generating Natural Adversarial Examples with Stable Diffusion." ICLR Tiny Papers (2024).
- Zhou, H., Lin, Y., et al. "BiSinger: Bilingual Singing Voice Synthesis." IEEE ASRU (2023).
References
[edit]- ^ "vLLM-Omni GitHub Repository". Retrieved 2026-03-24.
- ^ a b "Yueqian Lin profile". Scholars@Duke. Duke University. Retrieved 2026-03-24.
- ^ "DKU celebrates resilience and achievements of Class of 2024". Duke Kunshan University News. 2024-05-17. Retrieved 2026-03-24.
- ^ "Yueqian Lin at Commencement 2024". Duke Kunshan University News. 2024-05-20. Retrieved 2026-03-24.
- ^ a b c d e f "Yueqian Lin". Retrieved 2026-03-24.
- ^ Zhou, Huali; Lin, Yueqian (2023). "BiSinger: Bilingual Singing Voice Synthesis". arXiv:2309.14089.
{{cite arXiv}}: CS1 maint: missing class (link) A bot will complete this citation soon. Click here to jump the queue - ^ Shi, Jiatong; Lin, Yueqian. Singing Voice Data Scaling-up: An Introduction to ACE-Opencpop and KiSing-v2. Interspeech 2024.
- ^ Lin, Yueqian (2025). "HippoMM: Hippocampal-inspired Multimodal Memory for Long Audiovisual Event Understanding". arXiv:2504.10739.
{{cite arXiv}}: CS1 maint: missing class (link) A bot will complete this citation soon. Click here to jump the queue - ^ "SD-NAE: Generating Natural Adversarial Examples with Stable Diffusion". OpenReview. Retrieved 2026-03-24.
- ^ Wang, Zhongxi; Lin, Yueqian (2026). "MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Large Language Models". arXiv:2603.02482.
{{cite arXiv}}: CS1 maint: missing class (link) A bot will complete this citation soon. Click here to jump the queue
External links
[edit]- Official website
- Linyueqian publications indexed by Google Scholar
- Linyueqian on GitHub