Scholar
Miao Xiong
Google Scholar ID: yQ4U_5IAAAAJ
National University of Singapore
Citations & Impact
Citations
836
H-index
9
i10-index
9
Publications
20
Co-authors
8
list available
Publications
Resume (English only)
Academic Achievements
- Papers published:
- - "In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation" accepted at ICML 2024.
- - "Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs" accepted at ICLR 2024.
- - "Proximity-Informed Calibration of Deep Neural Networks" accepted at NeurIPS 2023 as a Spotlight.
- - "GraphCleaner: Detecting Mislabelled Samples in Popular Graph Learning Benchmarks" accepted at ICML 2023.
- - "Great Models Think Alike: Improving Model Reliability via Inter-Model Latent Agreement" accepted at ICML 2023.
- - "Birds of a Feather Trust Together: Knowing When to Trust a Classifier via Adaptive Neighborhood Aggregation" published in Transactions on Machine Learning Research (TMLR) 2022.
- - "Trust, but Verify: Using Self-supervised Probing to Improve Trustworthiness" accepted at ECCV 2022.
- - "Probabilistic Knowledge Distillation for Face Ensemble" accepted at CVPR 2023.
Research Experience
- Starting a new position as a Research Intern at Apple!
Education
- PhD student at National University of Singapore, advised by Prof. Bryan Hooi; Undergraduate with double majors in Computer Science and Statistics at Zhejiang University.
Background
- Research interests: Trustworthy and Responsible AI, especially addressing trust-related challenges in foundation models. Current research topics include LLMs (hallucination mitigation, RAG, constraint decoding, etc.). Also studies Uncertainty Estimation, in the context of Calibration, Failure Prediction, and Out-of-Distribution Detection to enhance the reliability of AI-based decision-making systems. Passionate about advancing AI towards greater safety, equity, and sustainability.
Miscellany
- Enthusiastic about embracing new challenges in this dynamic field.
