Pol Puigdemont

Efficient Transformer architectures

  • Spherical Fixed-Point Reasoners: Adaptive Loop Halting on the Hypersphere
    Arber Zela, Artavazd Maranjyan, Pol Puigdemont, David Berghaus, Valentyn Boreiko, Volkan Cevher.
    arXiv 2026[paper, post]
  • Spatial Priors via Space Filling Curves for Small and Limited Data Vision Transformers
    Leyla Naz Candogan, Arshia Afzal, Pol Puigdemont, Volkan Cevher.
    ICML 2026[paper, code]
  • Linear Attention for Efficient Bidirectional Sequence Modeling
    Arshia Afzal, Elias Abad Rocamora, Leyla Naz Candogan, Pol Puigdemont, Francesco Tonin, Yongtao Wu, Mahsa Shoaran, Volkan Cevher.
    NeurIPS 2025[paper, blog, code]
  • Probing hybridsWIP

AI Safety

  • Ascent Fails to Forget
    Pol Puigdemont*, Ioannis Mavrothalassitis*, Noam Levi*, Volkan Cevher.
    NeurIPS 2025[paper, post, talk]
  • Easy Data Unlearning Bench
    Pol Puigdemont*, Roy Rinberg*, Martin Pawelczyk, Volkan Cevher.
    MUGEN Workshop @ ICML 2025[paper, code, data]

Systems, Scheduling and Discrete Optimization

  • A Data-Driven Approach to Dataflow-Aware Online Scheduling for Graph Neural Network Inference
    Pol Puigdemont*, Enrico Russo*, Axel Wassington, Abhijit Das, Sergi Abadal, Maurizio Palesi.
    APSDAC 2025[paper]
  • Learning to Remove Cuts in Integer Linear Programming
    Pol Puigdemont*, Stratis Skoulakis*, Grigorios G Chrysos, Volkan Cevher.
    ICML 2024[paper]