Sonia Laguna
  • About Me
  • News
  • Outreach & Achievements
  • Publications
  • 📝 All Publications
    • Models Designed to Forget: Machine Unlearning via Key Deletion
    • Cartridges++: KV Cache Compression without Off-Context Derailment
    • ConRAG: Lightweight Inference of Multi-Hop Relations
    • Cyclostationary Phase Conditioning for Medical Time Series Diffusion
    • KV-Kaizen: Learning Context-Adaptive Cache Compression Choices
    • DynaMiCS: Fine-tuning LLMs with Performance Constraints using Dynamic Mixtures
    • Reference-Guided Machine Unlearning
    • Structure is Supervision: Multiview Masked Autoencoders for Radiology
    • Foundation Model for Cardiac Time Series via Masked Latent Attention
    • Post-hoc Stochastic Concept Bottleneck Models
    • Beyond Independent Frames: Latent Attention Masked Autoencoders for Multi-View Echocardiography
    • Concepts Based Methods
    • Steering Generative Models for Accessibility: EasyRead Image Generation
    • Interpretable Reward Modeling with Active Concept Bottlenecks
    • Leveraging the Structure of Medical Data for Improved Representation Learning
    • Beyond Glucose-Only Assessment: Advancing Nocturnal Hypoglycemia Prediction in Children with Type 1 Diabetes
    • Measuring Leakage in Concept-Based Methods: An Information Theoretic Approach
    • Text To 3D Object Generation For Scalable Room Assembly
    • Towards Scalable Newborn Screening: Automated General Movement Assessment in Uncontrolled Settings
    • Uncertainty Estimation for Trust Attribution to Speed-of-Sound Reconstruction with Variational Networks
    • RadVLM: A Multitask Conversational Vision-Language Model for Radiology
    • Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable
    • Stochastic Concept Bottleneck Models
    • Exploiting Interpretable Capabilities with Concept-Enhanced Diffusion and Prototype Networks
    • Synthetic data in generalizable, learning-based neuroimaging
    • Deep Generative Clustering With Multimodal Diffusion Variational Autoencoders
    • ExpLIMEable: An exploratory framework for LIME
    • Super-resolution of portable low-field MRI in real scenarios: integration with denoising and domain adaptation
    • Single Plane Illumination Microscopy for Microfluidic Device Imaging
    • Quantitative Brain Morphometry of Portable Low-Field-Strength MRI Using Super-Resolution Machine Learning
  • 🗞️ All News
    • 📝 MUNKEY (Machine Unlearning) got accepted at NeurIPS 2026🇦🇺!
    • 🖥️ Joined Apple as a Research Intern
    • 📝 Attending ICLR 2026 in Brazil, our five workshop papers got accepted with an oral contribution!
    • 📝 New paper on multimodal representation learning accepted to TMLR and on generative models for accessibility at CHI 2026
    • 🗣️ Invited Speaker at the Interpretable Deep Learning Reading Group and at the University of Zaragoza Seminar Series
    • 🗣️ Joined the organizing team of the ELLIS x UniReps Speaker Series
    • 🗣️ Speaker at the ETH-UN Forum and Selected Project at the ETH-UN Incubator
    • 📝 Summer research update! Four first author publications have been accepted and will be presented this summer at ICML, MLHC and IPCAI
    • 📝 Attending ICLR 2025 in Singapore, our four workshop papers got accepted with an oral (Best Paper Award 🏆) contribution!
    • 🗣️ Invited Speaker at the Microsoft Cambridge ML Seminar Series
    • 🖥️ Participated in the Spring into Quant 2025 G-Research Program!
    • 🏛️ Started a research stay at University of Cambridge!
    • 🗣️ I am Program Chair of the Synthetic Data Workshop at ICLR 2025 and of the GenAI for Biomed Applications Workshop at IEEE CBMS.
    • 🗣️ Our Workshop on GenAI meets XAI was accepted!
    • 📝 Our two main conference papers on Concept Bottleneck Models got accepted at NeurIPS 2024!
    • 📝 Our two workshop papers got accepted at NeurIPS 2024!
    • 🖥️ Joined Google as a Research Intern and later extended as a Student Researcher.
    • 📝 Our main conference paper on Deep Generative Clustering with Multimodal Diffusion Variational Autoencoders got accepted at ICLR 2024!
    • 📝 Our workshop paper got accepted at NeurIPS 2023.
    • 🗣️ Our Workshop on Deep Generative Models for Health was accepted!
    • 🏅 Featured in the Nova List top talents in Spain!
    • 📚 Attended the Machine Learning Summer School (MLSS) in South Africa!
    • 🏛️ Started my PhD in Machine Learning at ETH Zurich.
    • 🗣️ Joined CSNOW, 30th Anniversary and ETH Diversity Award.
    • 🏛️ Arrived to Boston for a research stay in Harvard!
    • 🏛️ Started the MSc in D-ITET at ETH Zurich!
    • 🏛️ Selected as Amgen Scholar to do research at ETH Zurich!
    • 🏛️ Started an exchange year at Georgia Institute of Technology!
    • 🏛️ Started my BSc in Biomedical Engineering at Universidad Carloss III de Madrid!
  • Blog
    • 🎉 First news
    • 🧠 Sharpen your thinking with a second brain
    • 📈 Communicate your results effectively with the best data visualizations
    • 👩🏼‍🏫 Teach academic courses
    • ✅ Manage your projects
  • 🗞️ All News
  • Publications
    • An example preprint / working paper
    • An example journal article
    • An example conference paper

Reference-Guided Machine Unlearning

Mar 14, 2026·
Jonas Mirlach*
Sonia Laguna*
Sonia Laguna*
,
Julia E Vogt
· 0 min read
PDF Code
Type
Conference paper
Publication
In ICLR 2026 Workshop AIWILD and extended Preprint - ArXiv
Last updated on Mar 14, 2026
Sonia Laguna*
Authors
Sonia Laguna*
Research @Apple MLR
PhD in Machine Learning @ETH Zurich
← DynaMiCS: Fine-tuning LLMs with Performance Constraints using Dynamic Mixtures Aug 20, 2026
Structure is Supervision: Multiview Masked Autoencoders for Radiology Mar 13, 2026 →

© 2026 Sonia Laguna. This work is licensed under CC BY NC ND 4.0

Published with Hugo Blox Builder — the free, open source website builder that empowers creators.