Scholar
Dylan Sam
Google Scholar ID: 43ffAwcAAAAJ
PhD Student, Carnegie Mellon University
Citations & Impact
Citations
158
H-index
7
i10-index
5
Publications
19
Co-authors
10
list available
Contact
Publications
Resume (English only)
Academic Achievements
- [May 2025] Gave a talk at UMD on 'Safety Pretraining', creating natively safe LLMs during pretraining
- [Feb 2025] Shared Google internship work on analyzing similarity metrics in LLM pretraining data curation
- [Jan 2025] Released new work on black-box monitoring of LLM performance and behaviors
- [Apr 2024] Presented 'Auditing Fairness under Unobserved Confounding' at AISTATS
- [Apr 2024] Presented work on generalization bounds for prompt engineering in VLMs at ICLR
- [Jul 2023] Gave a talk at the KLR workshop @ ICML 2023 on learning data-driven priors for BNNs incorporating interpretable domain knowledge
Research Experience
- Student Researcher at Google Research
- Research Intern at Amazon AWS
- Research Intern at Bosch Center for AI
- Research Intern at NASA JPL
- Summer 2024: Working at Google Research in NYC on data curation for LLM pretraining
Background
- Final-year PhD student in the Machine Learning Department at Carnegie Mellon University (CMU)
- Research focuses on making language models behave in safe, controllable, and predictable ways
- Current work includes curating safer training data and monitoring models for harmful behavior
- Previously worked on aligning models using weak supervision or interpretable domain knowledge, and on generalization
- Supported by an NSF Graduate Research Fellowship
- Currently collaborating with Gray Swan AI
