A Post-doc Researcher
in Computer Science
I am a post-doc researcher at the University of Copenhagen, and a member of CopeNLU and the Pioneer centre for artificial inteligence. My main research topics include multimodal fact-checking and misinformation detection, and explainable AI for NLP models.
I have obtained my PhD in Computer Science from the University of Copenhagen in 2025, supervised by Professor Isabelle Augenstein. My thesis, titled Revisiting Noise in Natural Language Processing for Computational Social Science, deals with the challenges (and opportunities) of working with noisy data in NLP for social science applications. I’ve completed an MSc and a BSs in Computer Science and Computational Biology at the Hebrew University of Jerusalem. I have also held the position of a Deep Learning tech lead at Lightricks.
I am passionate about helping social media users to better navigate the complex information ecosystem of today and be better equipped to identify misinformation and false narratives. I am also excited about creating methods to help both practitioners who use Neural Networks and the researchers who develop them to gain insights into why and how those models arrive at certain predictions. This, I believe, will increase the credibility and trust of ML tools in the public's view and promote the development of more capable and less biased models.
Here you can find my most recent publications and pre-prints.
We studied the relationship and interdependence between professional fact-checkers and Community Notes contributors in fact-checking social media posts.
Proceedings of ACL 2025
We trained a language model to extract moral values from over 20,000 subreddits. Our analysis reveals new insights into the relationship between human values and online communities.
Proceedings of NAACL 2025
We develop a novel methodology to extract political opinions and values from LLMs using tropes: semantically similar, recurrent phrases in the responses.
Findings of EMNLP 2024
We investigate the empirical ability of RNN and Transformer LMs to learn of regular, probabilistic languages with various complexity parameters.
Proceedings of ACL 2024
We study transformers' ability to learn random n-gram LMs to test whether theoretical lower bounds imply learnability of formal languages.
Proceedings of EMNLP 2024
This work presents Factcheck-Bench, a holistic end-to-end framework for annotating and evaluating the factuality of LLM-generated responses
Findings of EMNLP 2024
In this paper we explore an exciting alternative to the processing of historical documents with NLP tools – utilising OCR-free, pixel-based language models.
Proceedings of EMNLP 2023
We introduce a novel dataset of events extracted from multilingual historical newspapers, and use it to benchmark several LMs and training configurations.
Proceedings of ACL 2023
This study investigates intersectional biases in historical Caribbean newspapers, tracing their evolution and dynamics over a 250-year period.
Findings of ACL 2023
This paper presents a novel search engine based on a large-scale biological database to assist designers in finding biologically inspired solutions.
Proceedings of AAAI 2024
In this whimsical study, we curate and analyze a dataset of amusing scientific papers, developing NLP models to automatically identify them using hand-crafted features.
Proceedings of ACL 2021
I co-authored this paper while working at Lightricks, where I implemented the crowd-annotation platform and analysed the resulting data.
Proceedings of WACV 2022
Here you can download some of the posters, lectures and other media that I made and am proud of.
