
Germans Savcisens
PhD, Computational Social Science
Germans Savčišens
Герман Савчiшен
Герман Савчишен
Ghe-r-men Saf-chi-shen
- Veracity and Truth Tracking in LLMs
- Chains, Trees, and Graphs-of-Thought
- Beliefs in LLM–Human Networks
- Foundation Models for Event Data
Hi, I'm Germans Savcisens. I am a Postdoctoral Research Associate at the Khoury College of Computer Science at Northeastern University, where I am part of the RAD Lab and the Network Science Institute.
My research combines machine learning, network science, and human behavior. Right now, I am studying the knowledge and epistemic stability of large language models. I look at how LLMs embed knowledge, truth, and uncertainty, and what happens to those representations when models interact with each other.
I am originally from Latvia. Before moving to Boston, I earned my PhD from the Technical University of Denmark, where I worked on foundation models for longitudinal and behavioral data. This work led to the life2vec paper (Nature Computational Science), which introduces a transformer-based model that learns from individual life trajectories to predict health and socioeconomic outcomes.
I will be on the job market starting Spring 2027.
Latest Blog Posts
Dec 24, 2025
Is Life2vec a Mess?
What’s Wrong with AI Death Calculators?
A public-facing explanation of life2vec: its data, design, and limits, and why claims about '78% accuracy' and online death prediction are deeply misleading.