I'm a first-year CogSci PhD student at the Deep Cognition Lab at The Hebrew Universiry, advised by Dr. Ariel Goldstein. My work sits at the intersection of mechanistic interpretability and cognitive computational neuroscience.
My research focus is understanding and modeling the process of word selection in conversational speech with insights from LLMs.
Publications
pre-print 2025
The Road Not Taken: Unsaid Word Alternatives are Represented in the Brain
Daria Lioubashevski, Daniel Friedman, Adeen Flinker, Ariel Goldstein
For the 1st time we provide direct neural evidence for the simultaneous co-activation of multiple *unsaid* alternatives in the listener's and the speaker's brain during natural conversation.
ICML 2025
Looking Beyond The Top-1: Transformers Determine Top Tokens In Order
Daria Lioubashevski, Tomer Schlank, Gabriel Stanovsky, Ariel Goldstein
We show that Transformer models across text, vision and speech determine top ranking tokens *in order* of their ranking and uncover a novel task transition mechanism that explaind this phenomenon.
Education
Hebrew University of Jerusalem
PhD in Cognitive and Brain Sciences
Advisor: Dr. Ariel Goldstein
Hebrew University of Jerusalem
MSc in Computer Science
Advisor: Dr. Ariel Goldstein & Dr. Gabriel Stanovsky
Hebrew University of Jerusalem
BSc in Computer Science and Cognitive Science
Experience
Student Researcher — Google Research
Testing novel architectures for decoding language from brain signal, as part of a team applying deep learning methods to questions in computational neuroscience.
Computer Vision Algorithm Developer — Mobileye
Developed and deployed object detection modules for ADAS/AV under memory & realtime constraints combining CNNs and classical algorithms.