6 papers
From Mechanistic to Compositional Interpretability
Ward Gauderis, Thomas Dooms, Steven T. Homer +2
Mechanistic interpretability aims to explain neural model behaviour by reverse-engineering learned computational structure into human-understandable components. Without a formal fr…
NAACA: Training-Free NeuroAuditory Attentive Cognitive Architecture with Oscillatory Working Memory for Salience-Driven Attention Gating
Zhongju Yuan, Geraint Wiggins, Dick Botteldooren
Audio provides critical situational cues, yet current Audio Language Models (ALMs) face an attention bottleneck in long-form recordings where dominant background patterns can dilut…
Bilinear autoencoders find interpretable manifolds
Thomas Dooms, Ward Gauderis, Geraint Wiggins +1
Sparse autoencoders have become a standard tool for uncovering interpretable latent representations in neural networks. Yet salient concepts often span manifolds that current linea…
BioOSS: A Bio-Inspired Oscillatory State System with Spatio-Temporal Dynamics
Zhongju Yuan, Geraint Wiggins, Dick Botteldooren
Today's deep learning architectures are primarily based on perceptron models, which do not capture the oscillatory dynamics characteristic of biological neurons. Although oscillato…
Oscillatory Hierarchical Reservoirs for Human-like Rhythm Perception and Anticipation
Zhongju Yuan, Geraint Wiggins, Dick Botteldooren
Rhythm is a fundamental aspect of human behaviour, present from infancy and deeply embedded in cultural practices. Rhythm anticipation often occurs before surface event onsets, yet…
A General Close-loop Predictive Coding Framework for Auditory Working Memory
Zhongju Yuan, Geraint Wiggins, Dick Botteldooren
Auditory working memory is essential for various daily activities, such as language acquisition, conversation. It involves the temporary storage and manipulation of information tha…