Yuxin Wang
Scholar

Yuxin Wang

Google Scholar ID: T2S8PdUAAAAJ
Dartmouth College
NLPNatural Language UnderstandingKnowledge Representation and Reasoning
Citations & Impact
All-time
Citations
226
 
H-index
7
 
i10-index
7
 
Publications
11
 
Co-authors
0
 
Resume (English only)
Academic Achievements
  • Her work 'Backdooring VLMs via Concept-Driven Triggers' was accepted to ICML'25 DIG-BUG workshop. A preprint titled 'Probing Association Biases in LLM Moderation Over-Sensitivity' is under review. She has also published a paper titled 'ImpScore: A Learnable Metric For Quantifying The Implicitness Level of Sentences', which introduces a new method for quantifying the implicitness level of sentences.
Research Experience
  • She was a Research Intern at Microsoft Research Asia from July 2022 to February 2023, working on knowledge-enhanced open-world story generation. From June to September 2025, she interned at Microsoft Turing/MSR, focusing on conversational memory modeling and personalization of AI chatbot metrics. In September 2025, she will join Pinterest as a PhD Machine Learning Intern, where she will work on open-weight vLLMs inference optimization and model throughput acceleration.
Education
  • She obtained her bachelor's degree in Computer Science from Nanjing University in 2019 and her master's degree in 2023. During her master's, she conducted research on knowledge graph representation learning and continual learning at Websoft Lab.
Background
  • Yuxin is a third-year Ph.D. student in Computer Science at Dartmouth, advised by Prof. Soroush Vosoughi, with additional support from Prof. Saeed Hassanpour. Her research focuses on NLP, particularly the cognitive and societal aspects of AI. She primarily explores the cognitive and behavioral performances of AI systems by analyzing their language understanding capabilities and probing their reasoning processes. Additionally, she is interested in measuring and improving conversational AI’s integration with external tools (e.g., information retriever, strategy planner) to support more effective AI-human interaction.
Miscellany
  • She occasionally writes research and tech blogs on Medium and her personal blog. She enjoys jogging and cycling.
Co-authors
0 total
Co-authors: 0 (list not available)