Ayoub Hammal

I am a 2nd year PhD student in Natural Language Processing and Machine Learning.
I am working currently on Large Language Models' alignment and reasoning.
I mainly focus on developping test-time adaptation methods.
But my personal interest spans from convex optimization to formal languages.

Contact

  • <first name><family name><at>gmail<dot>com
  • My github

Education

  • PhD in Natural Language Processing 2022 - to this day
    LISN, Paris-Saclay University, CNRS.
    Supervised by Pr. Pierre Zweigenbaum, Dr. Caio Corro and Pr. Miguel Couceiro.
    I am mainly interested in this thesis in large language models' (LLMs) alignment.
    I focus my work on test-time alignment and domain adaptation approaches.
  • Master in Artificial Intelligence 2022 - 2024
    Paris-Saclay University. Valedictorian, Excellence honors.
  • Master 1 in Intelligent Computer Systems 2021 - 2022
    USTHB, Algiers. Valedictorian, Very Good honors.
  • Bachelor’s degree in Computer Science 2018 - 2021
    USTHB, Algiers. Valedictorian, Excellence honors.
  • Baccalaureate 2017 - 2018
    Mathematics option. Very Good honors.

Publications

Preprints

Publications

  • EuroBERT: Scaling Multilingual Encoders for European Languages
    Nicolas Boizard, Hippolyte Gisserot-Boukhlef, Duarte Miguel Alves, Andre Martins, Ayoub Hammal, Caio Corro, Celine Hudelot, Emmanuel Malherbe, Etienne Malaboeuf, Fanny Jourdan, Gabriel Hautreux, João Alves, Kevin El Haddad, Manuel Faysse, Maxime Peyrard, Nuno M Guerreiro, Patrick Fernandes, Ricardo Rei and Pierre Colombo.
    COLM 2025 - Second Conference on Language Modeling.
    Paper: https://openreview.net/forum?id=jdOC24msVq
  • Few-shot domain adaptation for named-entity recognition via joint constrained k-means and subspace selection
    Ayoub Hammal, Benno Uthayasooriyar and Caio Corro.
    COLING 2025 - Proceedings of the 31st International Conference on Computational Linguistics.
    Paper: https://aclanthology.org/2025.coling-main.662/
    Code: https://github.com/ayoubhammal/ckss4ner
  • Stable Matching based Resource Allocation for Large-Scale Fog Computing
    Khaled Zeraoulia, Ayoub Hammal, Mehdi Lerari and Youcef Hammal.
    IJCISIM 2023 - International Journal of Computer Information Systems and Industrial Management Applications 15.
  • An Efficient Resource Allocation Technique in a Fog Computing Environment
    Ayoub Hammal, Mehdi Lerari, Khaled Zeraoulia, and Youcef Hammal.
    ISDA 2022 - International Conference on Intelligent Systems and Digital Applications.

Academic work

  • Research Internship at ISIR, Sorbonne, Paris: Supervised and semi-supervised sequence-to-sequence models for structured prediction. 2024
    Supervised and semi-supervised sequence-to-sequence models for structured prediction.
    • Development of locally and globally normalized models for named-entity recognition task.
  • Research Work: Literature review on the mechanism of attention 2023
    Survey paper soon to be submitted for publication. I was interested mainly in:
    • The different approaches in the literature for modeling the soft/hard argmax operators intervening on the attention scores.
    • The different probing and analysis methods of trained attention-based networks, including its explainability and the linguistic features they entailed.
  • Research Internship at INRAE, AgroParisTech: Construction of a named-entities recognition system for out-of-vocabulary entities 2023
    • Implementation of state of the art NER models based on CLNER paper.
    • Proposal of an improved framework using enriched LLM-generated contexts.
  • Research Internship at INRAE, AgroParisTech: Enrich dietary data to make informed recommendations 2023
    • Construction of knowledge base around food consumption habits based on INCA2 survey.
    • Elaboration of a filtering system to validate food recommendations based on user profiles.
  • Research Internship at LISN, Paris Saclay: Study of generalization bound of loss functions and their consistency 2023
    • Study of generalization error of loss function and their different bounds.
    • Study of the Fisher Consistency property of loss functions.
    • Proposal of a new multiclass SVM loss function, and study of its Fisher Consistency property.
  • Research Internship at Laboratoire des Systèmes Informatiques (LSI), USTHB, Algiers: Design and implementation of a resource allocation technique for an IoT-Fog/Edge Computing environment 2021
    • I designed a resource allocation technique and algorithm for serving requests generated by IoT devices in a Fog/Edge Computing environment.
    • Then, I implemented and validated the technique in a simulation environment called IFogSim.
    • The algorithm is based on Gale-Shapley Matching algorithm, which I exploited and adapted to deploy requested services on heterogeneous Fog Nodes.

Scholarships

  • Eiffel Excellence scholarship 2022 - 2024
    For the Artificial Intelligence Master of the University of Paris-Saclay.

Thematic schools

Skills

  • Programming and scripting: Python, C/C++, Java, JS, Rust, Lua, Bash...
  • Scientific programming: Numpy, Matplotlib...
  • Deeplearning libraries: Pytorch, Transformers...
  • Distributed computing: Slurm, Jean Zay...