Seyed Arshan Dalili

Ph.D. Student in Computer Science and Engineering
image

Ph.D. student in Computer Science and Engineering at Pennsylvania State University, advised by Prof. Mehrdad Mahdavi (co-advised by Prof. Wenpeng Yin). Research spans mechanistic interpretability and representation learning for LLMs, federated learning and model merging, and hallucination mitigation. Collaborations include cancer multi-omics and Alzheimer’s biomarker discovery.


Advisor: Prof. Mehrdad Mahdavi | Co-advisor: Prof. Wenpeng Yin


 Curriculum Vitae

Education

  • Ph.D., Computer Science and Engineering
    Pennsylvania State University, State College, PA
    Jan 2025 - Present
  • B.Sc., Computer Engineering
    Sharif University of Technology, Tehran, Iran
    Sep 2019 - Jul 2023

Publications

Subspace-Aware Sparse Autoencoders for Mechanistic Interpretability (SASA)

Submitted to ICML 2026

S. A. Dalili, M. Mahdavi.

Model Merging via Multi-Teacher Knowledge Distillation

Submitted to ICML 2026 · arXiv:2512.21288

S. A. Dalili, M. Mahdavi.

Imaginations of WALL-E: Reconstructing Experiences with an Imagination-Inspired Module

arXiv:2308.10354

Z. S. Taghavi, S. Gooran, S. A. Dalili, H. Amirzadeh, M. J. Nematbakhsh, H. Sameti.

SUT at SemEval-2023 Task 1: Prompt Generation for Visual Word Sense Disambiguation

SemEval 2023 (ACL Anthology) · Paper

O. Ghahroodi, S. A. Dalili, S. Mesforoush, E. Asgari.

AIMA at SemEval-2024 Task 3: Simple Yet Powerful Emotion Cause Pair Analysis

SemEval 2024 (ACL Anthology) · Paper

A. Ghahramani Kure, M. Dehghani, M. M. Abootorabi, N. Ghazizadeh, S. A. Dalili, E. Asgari.

AIMA at SemEval-2024 Task 10: History-Based Emotion Recognition in Hindi-English Code-Mixed Conversations

SemEval 2024 (ACL Anthology) · Paper

M. M. Abootorabi, N. Ghazizadeh, S. A. Dalili, A. Ghahramani Kure, M. Dehghani, E. Asgari.

Research Experience

  • Graduate Research Assistant, Pennsylvania State University
    Advisor: Prof. Mehrdad Mahdavi. Research on mechanistic interpretability and representation learning for LLMs (e.g., sparse autoencoders), federated learning and model merging via distillation, and hallucination mitigation through inference-time steering. Collaborations on cancer multi-omics (with Prof. Jeffrey Maurice Peters) and Alzheimer's biomarkers (with Prof. Wenpeng Yin).
    Jan 2025 - Present
  • Research Assistant, UKP Lab, TU Darmstadt
    Supervisor: Prof. Iryna Gurevych. Implemented Fisher-weighted averaging and uncertainty-based gradient matching; contributed code to MergeKit.
    Aug 2024 - Nov 2024

Teaching Experience

CMPSC 448: Machine Learning and Algorithmic AI (Penn State)

Prof. Mehrdad Mahdavi | Spring 2025

Teaching Assistant, Sharif University of Technology

Courses: Artificial Intelligence (x3), Modern Information Retrieval, Computer Architecture, Advanced Programming, Circuits | 2020 - 2023

Working Experience

NLP Intern

ASR Gooyesh Pardaz | May 2023 - Jul 2023 | Tehran, Iran

Semantics and multimodal NLP; built large-scale IR models for Persian.

    Tutor, High School Mathematics and Physics

    Private tutoring | Jun 2020 - Feb 2021 | Sari, Iran

    One-on-one tutoring in mathematics and physics.

      Professional Service

      Academic Service

      • Reviewer, KDD 2025 Workshop: SciSocLLM.
      • Committee Member, Mid-Atlantic Student Colloquium on Speech, Language and Learning (MASC-SLL 2025).