Johannes Mario Meissner
Full-stack engineer and machine learning researcher based in Tokyo
Summary
Work
Education
Projects
Writing
Grad school in Japan: my experience doing a Master of Sciences in Computer Science
March 1, 2024Personal long-form article describing the author's experience in a research-focused master's program at the University of Tokyo, admissions, lab life, and publishing while enrolled.
Debiasing Masks: A New Framework for Shortcut Mitigation in NLU
January 1, 2022Proposes identifying debiased pruning masks that can be applied to finetuned models to mitigate shortcut behaviors in NLU, enabling selective, switchable debiasing without full retraining.
Embracing Ambiguity: Shifting the Training Target of NLI Models
January 1, 2021Explores training NLI models on annotator label distributions (ambiguity-aware targets) and introduces AmbiNLI to better capture linguistic ambiguity and improve representations.
Vulcanus in Japan: Arrival
February 1, 2020Blog post recounting the author's arrival and early days in the Vulcanus in Japan program, including language school and internship logistics.