Research Interests

I am interested in (1) rigorously understanding the capabilities and information inside machine learning models (see our work), and (2) building models that maximize reasoning capabilities, without relying on memorization (a cognitive core).

Publications

Operationalising the Superficial Alignment Hypothesis via Task Complexity
(In Review)
Tomás Vergara Browne, Darshan Patil, Ivan Titov, Siva Reddy, {Tiago Pimentel, Marius Mosbach}

Tracr-Injection: Distilling Algorithms into Pre-trained Language Models
ACL Findings 2025
Tomás Vergara Browne, Álvaro Soto

From Insights to Actions: The Impact of Interpretability and Analysis Research on NLP
EMNLP 2024
Marius Mosbach, {Vagrant Gautam, Tomás Vergara Browne}, Dietrich Klakow, Mor Geva

Eigenpruning (extended abstract)
LatinX Workshop @ NAACL 2024
Tomás Vergara Browne, Álvaro Soto, Akiko Aizawa

Large Language Models are biased to overestimate profoundness
EMNLP 2023
{Eugenio Herrera-Berg, Tomás Vergara Browne}, Pablo León-Villagrá, Marc Lluís-Vives, Cristian Buc Calderon

Talks

Interpretability and Analysis: An Overview and its Impact
Invited Talk @ AI Safety Initiative UC Chile

Teaching

Lecturer in ML for Computer Vision for the AI certificate of PUC
(December 2023 to December 2025)

Design and Analysis of Algorithms Teaching Assistant
(Spring 2023)

Linear Algebra Teaching Assistant (lectures in spanish)
(Fall 2020, Fall 2021)

Introduction to Programming Teaching Assistant
(Fall 2021)

Discrete Mathematics Teaching Assistant
(Spring 2020)

Other interests

Together with Benno Krojer we are starting a podcast, with a focus on shedding light on the untold stories of research in AI. This podcast is titled Behind the Research of AI.

I like reading Math, CS and Sci-fi books. You can check out some them here.

I really enjoy doing artistic gymnastics, although I am not great (but I'll get better over time).