Biology subjects

Nastase, S.

Publications and source records attributed to Nastase, S..

2 recordsLinked to original sources

bioRxiv2024

Meta-awareness, mind-wandering, and the control of 'default' external and internal orientations of attention

The "default mode" of cognition refers to the tendency to simulate internal experiences, rather than attending to external events in the moment. But in some contexts, external focus can become captivating enough to act as the default mode. To explore the relationship between prepotent internal and external default modes, we measured brain activity in forty participants using fMRI. Naturalistic movie clips were viewed, each one four times in sequence. When subjects were asked to focus attention on the videos, more mind-wandering events (distractions from the externally-focused task) occurred as the videos became less interesting with each repetition, and also when less engaging videos were presented. When subjects were asked to focus internally on breathing, more mind-wandering events (distractions from the internally-focused task) occurred when videos were most interesting (on the first repetition) and when more engaging videos were presented. In the fMRI data, inter-subject correlation, within-subject correlation, and GLM analyses found similar fronto-parietal networks engaged in transitions between default-controlled states regardless of the internal-external distinction, indicating more overlap in internal-external processing than previously assumed. We suggest that whether the default state is internal or external, and whether the sources that disrupt it are internal or external, depend on context.

neuroscience↗

bioRxiv2024

Scale matters: Large language models with billions (rather than millions) of parameters better match neural representations of natural language

Recent research has used large language models (LLMs) to study the neural basis of naturalistic language processing in the human brain. LLMs have rapidly grown in complexity, leading to improved language processing capabilities. However, neuroscience researchers havent kept up with the quick progress in LLM development. Here, we utilized several families of transformer-based LLMs to investigate the relationship between model size and their ability to capture linguistic information in the human brain. Crucially, a subset of LLMs were trained on a fixed training set, enabling us to dissociate model size from architecture and training set size. We used electrocorticography (ECoG) to measure neural activity in epilepsy patients while they listened to a 30-minute naturalistic audio story. We fit electrode-wise encoding models using contextual embeddings extracted from each hidden layer of the LLMs to predict word-level neural signals. In line with prior work, we found that larger LLMs better capture the structure of natural language and better predict neural activity. We also found a logarithmic relationship where the encoding performance peaks in relatively earlier layers as model size increases. We also observed variations in the best-performing layer across different brain regions, corresponding to an organized language processing hierarchy.

neuroscience↗