Through my research, I aim to develop ML models that can be trusted for high-stakes decision-making, with an emphasis on medical applications. I am particularly excited about questions related to causal reasoning, representation learning, …

Dec 9, 2024 · To overcome these limitations, we introduce Copyright-Protecting Model Fusion (CP-Fuse), a novel approach that combines models trained on disjoint sets of copyrighted material during inference.

Jan 22, 2025 · However, they often incur prohibitive computational costs or suffer from performance trade-offs. To overcome these limitations, we introduce Copyright-Protecting Model Fusion (CP-Fuse), a novel approach that combines models trained on disjoint sets of copyrighted material during inference.

Jul 29, 2024 · In this paper, we propose model fusion as an effective solution to safeguard against copyright infringement. In particular, we introduce Copyright-Protecting Fusion (CP-Fuse), an algorithm that adaptively combines language models …

To overcome these limitations, we introduce Copyright-Protecting Model Fusion (CP-Fuse), a novel approach that combines models trained on disjoint sets of copyrighted material during inference. In particular, CP-Fuse adaptively aggregates the model outputs to minimize the reproduction of copyrighted content, adhering

Javier Abad is a Doctoral Fellow at the ETH AI Center, supervised by Prof. Fanny Yang (Statistical Machine Learning group), Prof. Julia Vogt (Medical Data Science group), and Prof. Bjoern Menze (Department of Quantitative Biomedicine, University of Zurich).

Strong Copyright Protection for Language Models via Adaptive Model Fusion ( Poster ) > link Link Javier Abad · Konstantin Donhauser · Francesco Pinto · Fanny Yang

Jul 2, 2024 · In this paper, we propose model fusion as an effective solution to safeguard against copyright infringement. In particular, we introduce Copyright-Protecting Fusion (CP-Fuse), an algorithm that adaptively combines language models …

La lista denominada Citados por incluye las citas a los siguientes artículos de Google Académico. Los que se indican como * pueden diferir del artículo que aparece en el perfil.

Jun 26, 2023 · We prove that our method is a consistent approximation of full CP, and empirically show that the approximation error becomes smaller as the training set increases; e.g., for 1,000 training points the two methods output p-values that are <0.001 apart: a negligible error for any practical application.