Hi! I'm Sheza.

I'm a master's student at the University of Michigan, dedicated to advancing Responsible AI by enhancing the factuality of large language models and improving evaluation methods, with a particular focus on social NLP contexts.

I am actively looking for Fall '25 PhD positions in NLP and ML!

Publications

FactBench: A Dynamic Benchmark for In-the-Wild Language Model Factuality Evaluation

Farima Fatahi Bayat, Lechen Zhang, Sheza Munir, Lu Wang
Under Review at ICLR 2025

Deepfake Defense: Constructing and Evaluating a Specialized Urdu Deepfake Audio Dataset

Sheza Munir, Wassay Sajjad, Mukeet Raza, Emaan Abbas, Abdul Hameed Azeemi, Ihsan Ayyub Qazi, Agha Ali Raza
Association of Computational Linguistics (ACL) Findings 2024

Projects

  • LM Factuality Evaluation

    Developed VERIFY, a pipeline to evaluate LLMs' factuality, and created FACTBENCH a dynamic eval benchmark with 1000 prompts (Advisor: Dr. Lu Wang).

    Learn More
  • Deepfake Audio Defense Dataset

    Pioneered an Urdu speech training dataset for DeepFake detection with 35,000 audios, utilizing advanced end-to-end TTS models (Advisor: Dr. Agha Raza).

    Learn More
  • News-Economy Analysis

    Analyzed commodities data and news articles using Python, NLP, and regression analysis to identify economic correlations and forecast trends.

    Learn More
  • Text Style Transfer

    Developed a text style transfer system for classical-to-modern English using GPT-3.5-Turbo for data augmentation, creating a parallel dataset and training a BARTForConditionalGeneration model.

    Learn More
  • Migraine Prediction Engine

    Designed LSTM models on complex time-series health data from wearables and diaries, achieving 88% accuracy in predicting migraines one day in advance, while managing data pipelines.

    Learn More
  • Visual Instruction Tuning

    Reproducing Liu et al.'s work on LLaVA, a multimodal LLM trained on GPT-4-generated instruction data, focusing on improving performance in visual-language tasks.

    Project Underway

Teaching Experience

Teaching courses like Git, Shell, and Servers, Deep Learning, and Human-Computer Interaction has not only deepened my technical skills but also taught me to convey complex topics clearly and empathetically. Guiding students through challenging concepts has enhanced my adaptability and passion for creating an inclusive learning environment where both my students and I grow.

  • Deep Learning I
    SIADS 642
    UMich
    2023, 2024
  • Git, Shell, and Servers
    SI 504
    UMich
    2023, 2024
  • ML for Speech Processing
    CS 433
    LUMS
    2022
  • Human Computer Interaction
    CS 466
    LUMS
    2021
    500+
    Students Supported

Professional Experience

Research Assistant

LAnguage Understanding and generatioN researCH (LAUNCH) Lab

Jun. 2024 – Present

University of Michigan, Ann Arbor, MI

Graduate Student Instructor

SI 504-Git, Shell and Servers UMSI

Aug. 2023 – Present

University of Michigan, Ann Arbor, MI

OneTIS Intern

Trinity Health

May. 2024 – Dec. 2024

Ann Arbor, MI

Research Assistant - Lab Head

Center for Speech and Learning Technologies

Aug. 2021 – May 2023

Lahore University of Management Sciences, Lahore, Pakistan