About me

I am a forth-year Computer Science Ph.D. student at the UtahNLP Lab, University of Utah. My research explores LLM mechanistic interpretability, explainable AI, and applied NLP for social good. My work focuses on making AI systems transparent and reliable for real-world deployment. I develop methods that uncover how LLMs reason, evaluate when they might fail, and communicate uncertainty, all with an eye toward scalable, practical impact. This has led to multiple publications at top venues (including an Outstanding Paper nomination at EMNLP 2025), along with realistic, practical outputs like large-scale datasets and evaluation frameworks for agentic AI systems. I am advised by Dr. Ana Marasović.

I'm seeking research internship opportunities for Summer 2026. If my work aligns with your team's interests, I'd love to connect; feel free to reach out via email!

Daniel lewis
quote icon

Daniel lewis

Work in progress

  • Do Generated Chain-of-Thoughts Truly Mirror LLM Decision Making?
    Fall 2025


    Standing by your words
    [Image credit: DALL.E 3]

    Chain-of-thought is an effective method for tasks requiring multi-step reasoning, such as arithmetic and logic. Here, we aim to understand whether large language models genuinely use these thought chains for decision-making or if they are just auxiliary outputs. We propose a three-stage framework to mechanistically analyze the faithfulness of chain-of-thought circuits: constructing attribution graphs of model computations, extracting minimal computational backbones through graph skeletonization, and validating findings with targeted interventions. Combined with automated unfaithfulness detection, this approach enables systematic evaluation of whether step-by-step reasoning reflects genuine internal computations or merely serves as post-hoc justification.

Publications

  • Tutek, M., Hashemi Chaleshtori, Fateme, Marasovi´c, A., Belinkov, Y. S (Jul. 2025).
    “Measuring Faithfulness of Chains of Thought by Unlearning Reasoning Steps”.
    Empirical Methods in Natural Language Processing (EMNLP) 2025
    🏆 Nominated for Outstanding Paper Award at EMNLP 2025!
    🌟 Accepted to the Interplay workshop at COLM 2025 (non-archival)!
    Link to paper

  • Stringham, N., Hashemi Chaleshtori, Fateme, Yan, X., Xu, Z., Wang Phillips, B., Marasovi´c, A. (Oct. 2025).
    “Teaching People LLM’s Errors and Getting it Right”.
    Submitted to European Chapter of the Association for Computational Linguistics (EACL) 2025 (Under Review)

  • Hashemi Chaleshtori, Fateme, Ghosal, A., Gill, A., Bambroo, P., Ana Marasović, A. (Oct. 2024).
    “On Evaluating Explanation Utility for Human-AI Decision Making in NLP”.
    Findings of the Association for Computational Linguistics: EMNLP 2024
    🌟 Selected for oral presentation at XAI in Action Workshop at NeurIPS (non-archival)!
    Link to paper

  • Salek Faramarzi, N.*, Hashemi Chaleshtori, Fateme*, Shirazi, H., Banerjee, R., & Ray, I. (Apr. 2023).
    “Claim Extraction and Dynamic Stance Detection in COVID-19 Tweets”.
    Companion Proceedings of the ACM Web Conference (*equal contribution)
    Link to paper

  • Hashemi Chaleshtori, Fateme & Ray, I. (Dec. 2022).
    “Automation of Vulnerability Information Extraction Using Transformer-Based Models of Natural Language Processing”.
    European Symposium on Research in Computer Security (ESORICS 2022) Workshop on System Security Assurance (SecAssure 2022)
    Link to paper

  • Zou, C., Banerjee R., Hashemi Chaleshtori, Fateme Shirazi, H., & Ray, I. (Aug. 2021).
    “Seeing Should Probably not be Believing: The Role of Deceptive Support in COVID-19 Misinformation on Twitter”. .
    Journal of Data and Information Quality
    Link to paper

Resume

Research Interests

  1. Mechanistic Interpretability

  2. Explainable AI (XAI)

  3. Computing for Social Good

Education

  1. University of Utah

    2022 — 2026 (expected)

    Kahlert School of Computing - Computer Science - Ph.D.
    Advisor: Dr. Ana Marasović

  2. Colorado State University

    2019 — 2022

    Department of Computer Science - M.S.
    Thesis: COVID-19 Misinformation Detection in Twitter Through Authentic News Citation Detection
    Advisor: Dr. Indrakshi Ray
    GPA: 3.91/4

  3. Amirkabir University of Technology

    2014 — 2019

    Computer Eng. and I.T. Dept. - Computer Systems Architecture - B.Sc.
    Thesis: Parallelizing Anomaly Detection in Smart Home Activities
    Advisor: Dr. Maryam Amir Haeri
    GPA: 16.79/20 (3.48/4)

Honors and Awards

  1. Graduate Student Fellowship

    2022 - 2023

    School of Computing, University of Utah, Salt Lake City, UT

  2. Student Scholarship, ACM Richard Tapia Celebration of Diversity in Computing

    2022

    Washington, D.C.

  3. Ranked 4th with Parsian Robotics Group in RoboCup Competitions

    2017

    Small Size Soccer League, Nagoya, Japan

Services

  1. Reviewer

    2021 & 2022

    The Web Conference


    2022

    Conference on Data and Application Security and Privacy


    2021

    IEEE Transactions on Power Systems


    2021

    Australasian Conference on Information Security and Privacy

  2. President

    2021 - 2022

    Women in CyberSecurity chapter at Colorado State University

Teaching Experiences

  1. Mentor

    2021

    i-STEM Scholar for 1st-generation, low-income, and underrepresented young students


    2020 - 2021

    Ten undergraduate students working on NLP projects

  2. Teaching Assistant

    2024

    Natural Language Processing, University of Utah


    2021

    System Security, Colorado State University


    2020

    Introduction to Programming with Python, Colorado State University


    2019

    Parallel Programming, Colorado State University


    2015 & 2016

    Principles of Programming, Amirkabir University of Technology


    2016

    Data Structure and Algorithms, Amirkabir University of Technology