Salma Tarmoun

PhD student, University of Pennsylvania, University of Pennsylvania

Joined

September 2020

Names

Salma Tarmoun (Preferred)

Emails

****@jhu.edu (Confirmed)

****@sas.upenn.edu (Confirmed)

Personal Links

Google Scholar

DBLP

Career & Education History

PhD student

University of Pennsylvania, University of Pennsylvania (upenn.edu)

2023 – 2026

PhD student

Johns Hopkins University (jhu.edu)

2020 – 2023

Advisors, Relations & Conflicts

Coauthor

Ziqing Xu

Present

Postdoc Advisor

Lachlan Ewen MacDonald

Present

Coauthor

Hancheng Min

Present

PhD Advisor

Rene Vidal

2020 – Present

Expertise

Deep Learning

Present

Optimization

Present

Publications

Softmax-Induced Ill-Conditioning in Transformer Models
Salma Tarmoun, Lachlan Ewen MacDonald, Ziqing Xu, Hancheng Min, Rene Vidal
- Submitted to ICLR 2026
- Readers: Everyone
LoRA Provably Reduces Forgetting and Enables Adapter Merging in Multiclass Linear Classification
Ziqing Xu, Hancheng Min, Lachlan Ewen MacDonald, Salma Tarmoun, Weijie J Su, Enrique Mallada, Rene Vidal
- Submitted to ICLR 2026
- Readers: Everyone
Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares
Lachlan Ewen MacDonald, Hancheng Min, Leandro Palma, Salma Tarmoun, Ziqing Xu, Rene Vidal
- NeurIPS 2025 poster
- Readers: Everyone
A Local Polyak-Lojasiewicz and Descent Lemma of Gradient Descent For Overparametrized Linear Models
Ziqing Xu, Hancheng Min, Salma Tarmoun, Enrique Mallada, René Vidal
- CoRR 2025
- Readers: Everyone
Understanding the Learning Dynamics of LoRA: A Gradient Flow Perspective on Low-Rank Adaptation in Matrix Factorization
Ziqing Xu, Hancheng Min, Lachlan Ewen MacDonald, Jinqi Luo, Salma Tarmoun, Enrique Mallada, René Vidal
- CoRR 2025
- Readers: Everyone
A Local Polyak-Łojasiewicz and Descent Lemma of Gradient Descent For Overparametrized Linear Models
Ziqing Xu, Hancheng Min, Salma Tarmoun, Enrique Mallada, Rene Vidal
- Accepted by TMLR
- Readers: Everyone
A Local Polyak-Łojasiewicz and Descent Lemma of Gradient Descent For Overparametrized Linear Models
Ziqing Xu, Hancheng Min, Salma Tarmoun, Enrique Mallada, René Vidal
- Trans. Mach. Learn. Res. 2025
- Readers: Everyone
Understanding the Learning Dynamics of LoRA: A Gradient Flow Perspective on Low-Rank Adaptation in Matrix Factorization
Ziqing Xu, Hancheng Min, Lachlan Ewen MacDonald, Jinqi Luo, Salma Tarmoun, Enrique Mallada, Rene Vidal
- AISTATS 2025 Poster
- Readers: Everyone
Gradient Descent and Attention Models: Challenges Posed by the Softmax Function
Salma Tarmoun, Lachlan Ewen MacDonald, Hancheng Min, Ziqing Xu, Rene Vidal
- ICLR 2025 Conference Withdrawn Submission
- Readers: Everyone

View all 19 publications

Salma Tarmoun

Names

Emails

Personal Links

Career & Education History

Advisors, Relations & Conflicts

Expertise

Publications

Co-Authors