Arshia Afzal

prof_pic.jpg

EPFL STI IEM LIONS

ELD 244 (Bâtiment ELD)

Station 11, 1015 Lausanne

I’m Arshia, a third-year Ph.D. student in Electrical Engineering at EPFL advised by Professor Volkan Cevher. My research explores efficient and fast LLMs by merging deep learning architectures with systems.

Previously, I studied Electrical Engineering at Sharif University of Technology and earned a Bronze Medal at the 2018 International Physics Olympiad 🥉 and a Gold medal of National Physics Olympiad 🥇.

Beyond academics, I’m a passionate rock climber 🧗‍♂️ and pianist 🎹! :)

news

Mar 06, 2026 M Excited to join Mistral AI as a research intern! 🚀
Jan 26, 2026 ✍️ New blog post: On the Legacy of Linear Transformers as Positional Embeddings — exploring the duality of forget gates and position encodings! 📍
Jan 02, 2026 🏰 Selected for the PhD Summit at Coitadel! Honored to be among the participants. 🎓
Dec 01, 2025 🇧🇷 Our paper Selective Rotary Position Embedding got accepted at ICLR 2026! See you in Rio! 🎉
Sep 19, 2025 🦁 LION is accepted to NeuRIPS 2025!🍾

selected publications

  1. ICLR 2026
    srpe.png
    Selective Rotary Position Embedding
    Sajad Movahedi, Timur Carstensen, Arshia Afzal, and 3 more authors
    Dec 2025
  2. NeurIPS 2025
    lion.jpg
    Linear Attention for Efficient Bidirectional Sequence Modeling
    Arshia Afzal, Elias Abad Rocamora, Leyla Naz Candogan, and 5 more authors
    Dec 2025
  3. ICML 2024
    rest_eeg.png
    REST: Efficient and Accelerated EEG Seizure Analysis through Residual State Updates
    Arshia Afzal, Grigorios Chrysos, Volkan Cevher, and 1 more author
    , Vienna. Webpage of paper can be found here , Jul 2024