Yifei Ming
Contact: alvinming5 [at] gmail [dot] com

Hi! I am a research scientist at Salesforce AI Research. I am broadly interested in trustworthy AI that aligns with human values, especially in the era of foundation models. Topics that I currently focus on:
- Understand and improve reasoning of contextual LLMs and Agents.
- Obtain domain expert language models.
- Enable LLMs to handle long contexts effectively.
I obtained my CS Ph.D. in 2024 from the University of Wisconsin-Madison advised by Sharon Li. My thesis covers a spectrum of algorithmic developments for reliable foundation models in the open world. In the past, I have worked on Spatial Reasoning for LLMs and VLMs with Xin Wang, Vibhav Vineet, and Neel Joshi from Microsoft Research. I also worked on Multi-Modal Document OOD with Jiuxiang Gu from Adobe Research.
News
09/2024 | Is a picture worth a thousand words? 🤔 Check out our new work on spatial reasoning of LLMs and multi-modal LLMs at NeurIPS 2024 ✨ |
---|---|
05/2024 | Starting a new position as a Research Scientist at Salesforce! Super excited to explore the frontiers of LLM, VLM, reliable ML, among other fascinating topics. |
05/2024 |
Defended my Ph.D. thesis on Reliable Foundation Models in the Open World ![]() |
01/2024 |
Our paper Understanding Retrieval-Augmented Task Adaptation for Vision-Language Models is accepted by ICML 2024. ![]() |
01/2024 |
Our paper Provable Out-of-Distribution Generalization in Hypersphere is accepted by ICLR 2024. ![]() |
Publications
-
NeurIPSIs A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language ModelsIn Neural Information Processing Systems (NeurIPS) 2024
-
ICMLUnderstanding Retrieval-Augmented Task Adaptation for Vision-Language ModelsIn International Conference on Machine Learning (ICML) 2024
-
ICLRProvable Out-of-Distribution Generalization in HypersphereIn International Conference on Learning Representations (ICLR) 2024
-
CPAL
Oral Domain Generalization via Nuclear Norm RegularizationIn Conference on Parsimony and Learning (CPAL) 2023 -
EMNLPA Critical Analysis of Document Out-of-Distribution DetectionIn Empirical Methods in Natural Language Processing (EMNLP Findings) 2023
-
IJCVHow Does Fine-Tuning Impact Out-of-Distribution Detection for Vision-Language Models?In International Journal of Computer Vision (IJCV) 2023
-
NeurIPSDomain Generalization with Nuclear Norm RegularizationIn Neural Information Processing Systems (NeurIPS’W) DistShift Workshop 2022
-
ICMLAre Vision Transformers Robust to Spurious Correlations?In International Conference on Machine Learning (ICML’W), SCIS Workshop 2022