Samuel Joseph Amouyal

Samuel Joseph Amouyal

PhD candidate @ TAU, Research scientist @ LegionSecurity

TAU-NLP

LegionSecurity

Biography

About me

I am a PhD student in NLP under the supervision of the amazing Prof. Jonathan Berant and Prof. Aya Meltzer-Asscher at Tel-Aviv University in Israel. I mainly research Large Language Models (LLMs) under the scope of psycholinguistic theory, hoping to draw conclusions for both NLP and psycholinguistics.
I am also a research scientist at LegionSecurity, working on evaluations and web-agent research. Before that I was the NLP research lead at BlinqIO.

Additional research interests

I began my academic journey in the Adi Lautman Interdisciplinary Program for Outstanding Students and I took classes in many faculties (mainly neurology, psychology, philosophy, biology but also economics, history, education…).
My research interest are therefore diverse and I have a few ongoing projects apart from my main PhD research:

  • A project about consciousness and LLMs
  • Projects in economy/game theory
    I am always open to new collaborations and exploring new fields so feel free to reach out.
Interests
  • NLP
  • Brain science
  • Literature
Education
  • PhD in Computer Science (NLP), ongoing, 2022

    Tel-Aviv University (Israel)

  • MSc in Computer Science (NLP), 2022

    Tel-Aviv University (Israel)

  • Adi Lautman Interdisciplinary Program for Outstanding Students, 2018

    Tel-Aviv University (Israel)

Recent Publications

(2025). Comparing Human and Language Models Sentence Processing Difficulties on Complex Structures. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), and the 39th Annual Conference on Human Sentence Processing (HSP 2026).

Cite Link Code

(2025). Partial reanalysis: A large-scale replication testing plausibility and transitivity bias. 38th Annual Conference on Human Sentence Processing (HSP), College Park, Maryland.

Cite Abstract

(2025). When the LM misunderstood the human chuckled Analyzing garden path effects in humans and language models. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics.

Cite Link arXiv Code

(2024). GLEE: A Unified Framework and Benchmark for Language-based Economic Environments. SEA Workshop, Conference on Neural Information Processing Systems (NeurIPS).

Cite Link Code

(2024). AssistantBench Can Web Agents Solve Realistic and Time-Consuming Tasks?. Proceedings of the Conference on Empirical Methods in Natural Language Processing.

Cite Link Code

Experience

 
 
 
 
 
Research scientist
May 2025 – Present Tel-Aviv, Israel
Research scientist at LegionSecurity, a startup developing an AI agent that performs SOC analyst tasks. Working on evaluations and web-agent research.
 
 
 
 
 
NLP research lead
November 2023 – April 2025 Tel-Aviv, Israel
NLP research lead for BlinqIO, a startup developing a generative AI agent that replaces human QA engineers. Led the company NLP research across training, prompt engineering, and pre/post-processing algorithms.
 
 
 
 
 
Teaching Assistant, Computational Models
October 2022 – Present Tel-Aviv, Israel
Teach tutorials for the Computational Models course to classes of 50-60 students.
 
 
 
 
 
Research scientist
July 2022 – November 2023 Tel-Aviv, Israel
Research scientist in the GT2D R&D team, leading research on two projects: allowing multi-modal models to reason over images, and model selection using deep learning. Built the deep learning training library currently used by the GT team (200 people).
 
 
 
 
 
Algorithm engineer
July 2020 – June 2022 Tel-Aviv, Israel
Part of the GT2D team, responsible for creating algorithms (models and post-processing) that extract the ground truth used across Mobileye from 2D images.

Awards & Honors

Vatat Grant for Outstanding PhD Students in Data Science
TAD Grant
Adi Lautman Excellence Scholarship

Contact

Let’s keep in touch if you have any questions about my research or want to talk about potential collaboration!