I am a fourth-year PhD candidate at the University of Michigan in Ann Arbor and an intern at Ai2. I am advised by Lu Wang and Honglak Lee. My main research interests are Large Language Models, Reasoning, and Controlled Generation. I spent 10 months at Amazon AI working with Miguel Ballesteros and Kathy Mckeown on multiple projects including Dialogue Summarization and Semi-structured documents understanding. Prior to that, I was an intern at Naver Labs Europe where I worked on Controllable Text Generation and Energy-based models with Hady Elsahar and Marc Dymetman.
On a side note, I’m an avid reader of psychology, and philosophy, and an all-time listener of uplifting trance music. In my free time, I play the piano, write, and produce my own music. If you’d like to chat about research, mentoring, or potentially collaborate, ping me at [lastname]m@umich.edu.
News
April 2025 Our work on Generative Process Reward Models is out!.
April 2025 Was awarded the CSE Service Award at UMich! 🎉
April 2025 Gave a talk on model merging at DLCT. Here are the Slides
March 2025 Co-organized the NLP@Michigan Day
December 2024 My work at Cohere on large-scale model merging optimization is out!
July 2024: Two papers accepted at COLM and ACL Findings!
June 2024: Started my internship at Cohere!
April 2024: My work with Ai2 on on source-aware training is released on arxiv
March, 2024: Gave a talk on LLM Reasoning at ML Collective DLCT reading group. Video is here.
Nov, 2023: Gave an invited talk at NAVER Labs Europe on Discriminator-Guided Chain-of-Thought. Slides are here.
Oct, 2023: Two papers accepted at EMNLP 2023 (Findings and main)!
May, 2023: Three papers accepted at ACL 2023 (two at the main conference and one at Findings)!
Nov 30th, 2022: I was awarded the Rackham Fellowship for Outstanding International Students!
Highlighted Papers
(Likely outdated; For an up-to-date list, visit my google scholar)
Muhammad Khalifa, Lajanugen Logeswaran, Moontae Lee, Honglak Lee, Lu Wang. Exploring Demonstration Ensembling for In-context Learning. ICLR 2023 Workshop on Mathematical and Empirical Understanding of Foundation Models. [Paper].
Muhammad Khalifa, Lajanugen Logeswaran, Moontae Lee, Honglak Lee, Lu Wang. Few-shot Reranking for Multi-hop QA via Language Model Prompting. ACL 2023 [Paper] [Code].
Muhammad Khalifa, Miguel Ballesteros, and Kathleen McKeown. 2021. A Bag of Tricks for Dialogue Summarization. EMNLP 2021 (short paper) [Paper].
Muhammad Khalifa, Muhammad Abdul-Mageed, Khaled Shaalan. “Self-Training Pre-Trained Language Models for Zero-and Few-Shot Multi-Dialectal Arabic Sequence Labeling.” EACL 2021. [Paper] [Code] [Bibtex]
Muhammad Khalifa*, Hady Elsahar*, Marc Dymetman*. “A Distributional Approach to Controlled Text Generation”. ICLR 2021. [Paper] [Code] [Blog]