Mehrdad Farajtabar

Researcher, Apple

Joined

September 2018

Names

Mehrdad Farajtabar (Preferred)

Emails

****@gatech.edu (Confirmed)

****@gmail.com (Confirmed)

****@google.com (Confirmed)

****@apple.com (Confirmed)

Personal Links

Career & Education History

Researcher

Apple (apple.com)

2022 – Present

Research Scientist

Google (google.com)

2018 – 2022

PhD student

Georgia Institute of Technology (gatech.edu)

2013 – 2018

Advisors, Relations & Conflicts

PhD Advisor

Le Song

2014 – 2018

PhD Advisor

Hongyuan Zha

2013 – 2018

Expertise

LLM Reasoning

Present

LLM Efficient Inference

Present

Mixture of Expert Models

Present

Large Language Models

Present

Clip Models

2022 – Present

Multimodal Models

2022 – Present

Continual Learning

2019 – Present

Publications

RL for Reasoning by Adaptively Revealing Rationales
Mohammad Hossein Amani, Aryo Lotfi, Nicolas Baldwin, Samy Bengio, Mehrdad Farajtabar, Emmanuel Abbe, Robert West
- ICLR 2026 Poster
- Readers: Everyone
Barriers for Learning in an Evolving World: Mathematical Understanding of Loss of Plasticity
Amir Joudaki, Giulia Lanzillotta, Mohammad Samragh, Seyed Iman Mirzadeh, Keivan Alizadeh, Thomas Hofmann, Mehrdad Farajtabar, Fartash Faghri
- ICLR 2026 Poster
- Readers: Everyone
MoE-PHDS: One MoE checkpoint for flexible runtime sparsity
Lauren Hannah, Soheil Zibakhsh Shabgahi, Kumari Nishu, Arnav Kundu, Mehrdad Farajtabar, Minsik Cho, Mohammad Samragh
- Submitted to ICLR 2026
- Readers: Everyone
The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity
Parshin Shojaee, Seyed Iman Mirzadeh, Keivan Alizadeh, Maxwell Horton, Samy Bengio, Mehrdad Farajtabar
- NeurIPS 2025 poster
- Readers: Everyone
From Dense to Dynamic: Token-Difficulty Driven MoEfication of Pre-Trained LLMs
Kumari Nishu, Sachin Mehta, Samira Abnar, Mehrdad Farajtabar, Maxwell Horton, Mahyar Najibi, Moin Nabi, Minsik Cho, Devang Naik
- SLLM
- Readers: Everyone
Proxy-FDA: Proxy-based Feature Distribution Alignment for Fine-tuning Vision Foundation Models without Forgetting
Chen Huang, Skyler Seto, Hadi Pouransari, Mehrdad Farajtabar, Raviteja Vemulapalli, Fartash Faghri, Oncel Tuzel, Barry-John Theobald, Joshua M. Susskind
- ICML 2025 poster
- Readers: Everyone
Model Soup for Better RLHF: Weight Space Averaging to Improve Alignment in LLMs
Atoosa Chegini, Hamid Kazemi, Seyed Iman Mirzadeh, Dong Yin, Maxwell Horton, Moin Nabi, Mehrdad Farajtabar, Keivan Alizadeh
- FITML 2024 Poster
- Readers: Everyone
GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models
Seyed Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, Mehrdad Farajtabar
- ICLR 2025 Poster
- Readers: Everyone
TiC-LM: A Multi-Year Benchmark for Continual Pretraining of Language Models
Jeffrey Li, Mohammadreza Armandpour, Seyed Iman Mirzadeh, Sachin Mehta, Vaishaal Shankar, Raviteja Vemulapalli, Samy Bengio, Oncel Tuzel, Mehrdad Farajtabar, Hadi Pouransari, Fartash Faghri
- Submitted to ICLR 2025
- Readers: Everyone
Proxy-FDA: Proxy-based Feature Distribution Alignment for Fine-tuning Vision Foundation Models without Forgetting
Chen Huang, Skyler Seto, Hadi Pouransari, Mehrdad Farajtabar, Raviteja Vemulapalli, Fartash Faghri, Oncel Tuzel, Barry-John Theobald, Joshua M. Susskind
- Submitted to ICLR 2025
- Readers: Everyone

View all 45 publications

Co-Authors

View all 100 co-authors