7 papers
Shared Circuits for Shared Grammar: Tracing Subject-Verb Agreement Across Languages
Isabella Gidi, Antonio Almudévar, Core Francisco Park +2
Multilingual large language models often generalize across languages, and prior work suggests that their internal mechanisms can overlap cross-lingually. It remains unclear, howeve…
Open-Set Source Tracing as Compositional Factors via Structured Prototypes
Santiago Rubio, Antonio Almudévar, Antonio Miguel +2
Recent research expands beyond binary anti-spoofing with the emergence of Source Tracing, the task of identifying the specific generative origins of synthetic speech. However, curr…
Rethinking Disentanglement under Dependent Factors of Variation
Antonio Almudévar, Alfonso Ortega
Representation learning is an approach that allows to discover and extract the factors of variation from the data. Intuitively, a representation is said to be disentangled if it se…
There Was Never a Bottleneck in Concept Bottleneck Models
Antonio Almudévar, José Miguel Hernández-Lobato, Alfonso Ortega
Deep learning representations are often difficult to interpret, which can hinder their deployment in sensitive applications. Concept Bottleneck Models (CBMs) have emerged as a prom…
Sparse Autoencoders Make Audio Foundation Models more Explainable
Théo Mariotte, Martin Lebourdais, Antonio Almudévar +3
Audio pretrained models are widely employed to solve various tasks in speech processing, sound event detection, or music information retrieval. However, the representations learned…
Aligning Multimodal Representations through an Information Bottleneck
Antonio Almudévar, José Miguel Hernández-Lobato, Sameer Khurana +2
Contrastive losses have been extensively used as a tool for multimodal representation learning. However, it has been empirically observed that their use is not effective to learn a…