Donato Crisostomi

Donato Crisostomi

ELLIS Ph.D. Student @ Sapienza/Cambridge

Biography

I am a 3rd year Ph.D. student in Computer Science at Sapienza University of Rome, where I am part of the GLADIA research group, led by prof. Emanuele Rodolà.

I believe curiosity to have a primary role in guiding research, and hence I am always looking forward to explore new directions in my field. My background includes hands-on research experience in Natural Language Understanding, Computer Vision and Geometric Deep Learning. Some of my more stable interests include Model Merging and Representational Aligment.

I wholeheartedly advocate for cleaner code in ML, as complexity should not be fought with more complexity. Maybe unsurprisingly at this point, I really enjoy coding. Funnily enough, I enjoy it so much that I often code as a mean to get away from a tiring day of coding.

In my (too little) free time, I enjoy throwing away my laptop to reconnect with nature through hiking and/or camping. Finally, like most of the humans on this planet, I enjoy travelling and good food.

Education
  • Ph.D. in Computer Science, present

    ELLIS PhD @ Sapienza University & University of Cambridge

  • MS.c. in Computer Science, 2021

    Sapienza University

    110/110 with honours | GPA 30.5/31

  • BS.c. in Computer Science, 2019

    Sapienza University

    110/110 with honours | GPA 29.6/31

Experience
  • Applied Scientist

    Jun 2022 - Dec 2022

    Amazon Search

  • Research Scientist

    May 2021 - Oct 2021

    Amazon Alexa

News

Attended the amazing EEML 2025 and won a best poster award! 🏅 Check it here
Mentored a great team at LOGML 2025 and helped them achieve the best project award! 🏆 Check it here 👀
Our paper “Mergenetic: a Simple Evolutionary Model Merging Library” has been accepted at ACL 2025, System Demonstration Track. Come evolve your own LLMs on Mergenetic 🧬!
Both MERGE3 and Update Your Transformer were accepted at ICML 2025 main track! 🙏
Check our new preprint MASS to see MoEs and model merging in action 💥
Audio pre-prints 🎧! Check these out to see stemmed accompaniment generation, how to generate seamless loops and activation patching for steering music generation.
The project NEXUS was awarded 1.5M€ by the Italian Fund for Science 🚀 I will be leading the Model Reuse work package!
Our Task Singular Vectors paper has been accepted at CVPR 2025! Check out the twee-print here.
It’s pre-print time! Check these out to see evolutionary model merging on a 4090 and the relation between task vectors and gradients.
Our Cycle-Consistent Multi-Model Merging paper has been accepted at NeurIPS 2024! 🎉