22分钟前更新- 01Reward Valuation in Vision Language Models: Causal Mechanisms Underlying AnhedoniaarXiv:2607.06626v1 Announce Type: cross Abstract: Recent Vision-Language Models capture increasingly complex aspects of human cognition. Here we ask whether this alignment extends to reward valuation, which we assess in a mechanistic framework built on clinical tests that were developed to evaluate anhedonia and motivational deficits in major depressive disorder. In the brain, anhedonia is frequently linked to dysregulation in the Nucleus Accumbens (NAc) and the broader dopaminergic reward systeMelika Honarmand, Samin Mahdipour Aghabagher, Martin Schrimpf
- 02STST-JEPA: Shallow-Target Spatio-Temporal Joint Embedding Prediction Architecture For EEG Self-Supervised LearningarXiv:2607.06629v1 Announce Type: cross Abstract: Brain age -- the age inferred from a physiological recording -- is an emerging biomarker whose deviation from chronological age tracks neurological and psychiatric burden, and EEG is an attractive substrate for it because it is cheap, portable, and temporally rich. Yet EEG brain-age models must contend with cross-site montage heterogeneity, small labelled cohorts, and dominant subject-level non-stationarity, and few EEG foundation models have beeRoy Segal, Yoni Svechinsky, Tomer Fekete
- 03Single-Entity Spiking Neuron Models: SurveyarXiv:2607.07429v1 Announce Type: cross Abstract: In this work, we reviewed different approaches in mathematical modeling of biologically plausible neural systems. Models are characterized and classified based on their common features and special use cases. In addition to spiking models, different types of discrete and continuous analogs are considered to accurately simulate biological processes, including membrane potential dynamics. The models under investigation include neurons and various coLeon Parepko, Danila Shulepin, Albert Nasybullin