Hi! My name is Jing Ding (丁婧 in Chinese). I’m a senior undergraduate student majoring in Data Science and Mathematics at the University of Michigan, where I am part of the SLED Lab. My research broadly focuses on multimodal AI, particularly vision-language models, interpretability, and embodied AI. I am also interested in cognitive science and psycholinguistics. Before coming to Michigan, I studied at Huazhong University of Science and Technology.
Always excited to connect, collaborate, and share ideas—feel free to reach out!
🔥 News
- 2026.09: Heading to IROS 2026 - see you in Pittsburgh!
- 2026.09: “SIMLIFE: Pattern Understanding for Long-Horizon Human-Agent Partnership” was accepted to LSEI@COLM 2026 and selected as a lightening talk!
- 2026.08: “Metonymic Circuits for Abstract Concept Grounding in Vision Transformers” was accepted to EMNLP 2026 as a main conference paper! See you online!
- 2025.07: “Vision-Language Models Are Not Pragmatically Competent in Referring Expression Generation” was accepted to COLM 2025! See you in Montreal!
- 2025.06: Our paper “Vision-Language Models Are Not Pragmatically Competent in Referring Expression Generation” was accepted to GEM@ACL 2025 and selected as a Spotlight at CVinW@CVPR 2025! See you in Nashville!
- 2025.05: Heading to NAACL 2025 — hope to see you there!
- 2025.04: Glad to presented at the first CSE Undergraduate Research Symposium!
- 2025.04: 🎉🎉 Our work “Vision-Language Models Are Not Pragmatically Competent in Referring Expression Generation” is on arXiv!
📝 Publications
Metonymic Circuits for Abstract Concept Grounding in Vision Transformers
Jing Ding, Ziqiao Ma, Jiayuan Mao, Joyce Chai, Freda Shi
TL;DR: Using Transcoder-based circuit tracing on a curated icon dataset, we show that vision transformers ground abstract concepts through causally important concrete visual anchors.
SIMLIFE: Pattern Understanding for Long-Horizon Human-Agent Partnership
Lightening talk at LSEI@COLM 2026
Run Peng, Zinnia Nie, Jing Ding, Yinpei Dai, Yichi Zhang, Zengqing Wu, Yao Fu, Ziqiao Ma, Jiayuan Mao, Joyce Chai
TL;DR: We study how AI models understand patterns through ultra-long term observations.
Vision-Language Models Are Not Pragmatically Competent in Referring Expression Generation
GEM@ACL 2025; Spotlight at CVinW@CVPR 2025
Ziqiao Ma*, Jing Ding*, Xuejun Zhang, Dezhi Luo, Jiahe Ding, Sihan Xu, Yuchen Huang, Run Peng, Joyce Chai
Paper | Homepage | Code | Dataset
TL;DR: We show significant pragmatic deficiencies in current VLMs when faced with referring expression generation compared to humans, as they violate Gricean maxims.
📖 Educations
- 2025.01 - 2026.12 (now), Honor Data Science & Mathematics, LSA, University of Michigan, Ann Arbor
- 2022.09 - 2024.06, Electronic Information and Engineering, School of Electronic Information and Communications, Huazhong University of Science and Technology
💻 Internships
- 2024.07 - 2024.08, iSURE Program, University of Notre Dame, USA
🌎 Visitor Map
© 2025 Jing Ding
