Scholar
Bangzhao Shu
Google Scholar ID: 3VdLBgoAAAAJ
Northeastern University
Citations & Impact
Citations
55
H-index
1
i10-index
1
Publications
3
Co-authors
5
list available
Publications
Resume (English only)
Academic Achievements
- 1. Paper 'Fluent but Unfeeling: The Emotional Blind Spots of Language Models' accepted to ICWSM 2026, which benchmarks a fine-grained emotion recognition dataset EXPRESS with 251 unique emotion labels.
- 2. Paper 'Causally Modeling the Linguistic and Social Factors that Predict Email Response' accepted to NAACL 2025.
- 3. Paper 'You don’t need a personality test to know these models are unreliable: Assessing the Reliability of Large Language Models on Psychometric Instruments' accepted to NAACL 2024, constructs a dataset covering 39 psychometric instruments across 115 persona axes, showing current question-style prompts are insufficient to capture model perceptions reliably.
Research Experience
- 1. Teaching Assistant for CS 6120 Natural Language Processing, Northeastern University, Winter 2025
Education
- 1. Ph.D. in Computer Science, Northeastern University, Expected 2024–2029, Advisor: Mai ElSherief
- 2. M.S. in Information, University of Michigan, 2022–2024, Advisor: David Jurgens
- 3. M.S. in Environment & Sustainability, University of Michigan, 2021–2024, Advisor: Yuhao Kang (University of Texas, Austin)
- 4. B.E. in Geomatics, Nanjing Normal University, 2016–2020
Background
- Broad interests in Natural Language Processing, Machine Learning, and Large Language Models. Focuses on developing robust and trustworthy language models by evaluating and improving AI safety and alignment through benchmarking, training, mechanistic interpretability, and steering. Also has experience in affective computing, model personas, and computational social science.
