Federico Bartsch

 Trento, Italy

 University of Trento

Federico Bartsch

PhD Researcher · MHUG Group · University of Trento

I am a PhD student in Artificial Intelligence at the University of Trento, supervised by Elisa Ricci and Massimiliano Mancini in the Multimedia and Human Understanding Group (MHUG). My research focuses on vision-language models.

news

Sep 2024 Started research internship at Toyota Motor Europe, Brussels.
Oct 2025 Started PhD in Artificial Intelligence at the University of Trento.
May 2026 Paper accepted at CVPR 2026 Findings — SEM: Sparse Embedding Modulation for Post-Hoc Debiasing.

publications

  1. CVPR 2026

    SEM: Sparse Embedding Modulation for Post-Hoc Debiasing

    Quentin Guimard*, Federico Bartsch*, Simone Caldarella, Rahaf Aljundi, Elisa Ricci, Massimiliano Mancini  · *Equal contribution

    CVPR Findings, 2026

    A zero-shot framework using Sparse Autoencoders (SAEs) to decompose CLIP text embeddings into interpretable, disentangled features. A modulation mechanism isolates and neutralizes bias-relevant neurons while preserving query-relevant semantic fidelity in the latent space.

    PDF arXiv

cv

Education

2025 – present PhD in Artificial Intelligence
University of Trento · MHUG Group
2023 – 2025 MSc in Artificial Intelligence Systems
University of Trento
2019 – 2023 BSc in Computer Science
University of Milan Bicocca

Experience

09/2024 – 03/2025 Toyota Motor Europe — Brussels, Belgium
Computer Vision Researcher Intern · Supervisor: Wim Abbeloos
  • Researched high-resolution image utilization for semantic segmentation and object detection.
  • Developed Early Exit methods for segmentation, reducing inference time by 40% with under 1% accuracy loss.
10/2022 – 04/2023 Hewlett Packard Enterprise — Rome, Italy
Software Engineer Intern (Machine Learning)
  • Built an anomaly detection system for multivariate time series data from 5G antennas.