Language models have been shown to be very effective in predicting brain recordings of subjects experiencing complex language stimuli.

However,itisstill not understood howthe task itself contributes to this difference.

Language models have been shown to be very effective in predicting brain recordings of subjects experiencing complex language stimuli. For a deeper understanding of this alignment, it is important to understand the correspondence between the detailed processing of linguistic information by the human brain versus language models. We investigate this correspondence via a direct approach, in which we eliminate information related to specific linguistic properties in the language model representations and observe how this intervention affects the alignment with fMRI brain recordings obtained while participants listened to a story. We investigate a range of linguistic properties (surface, syntactic, and semantic) and find that the elimination of each one results in a significant decrease in brain alignment. Specifically, we find that syntactic properties (i.e. Top Constituents and Tree Depth) have the largest effect on the trend of brain alignment across model layers. These findings provide clear evidence for the role of specific linguistic information in the alignment between brain and language models, and open new avenues for mapping the joint information processing in both systems.

What aspects of computer programs are represented by the human brain during comprehension? We leverage brain recordings derived from functional magnetic resonance imaging (fMRI) studies of programmers comprehending Python code to evaluate the properties and code-related information encoded in the neural signal. We first evaluate a selection of static and dynamic code properties, such as abstract syntax tree (AST)-related and runtime-related metrics. Then, to learn whether brain representations encode fine-grained information about computer programs, we train a probe to align brain recordings with representations learned by a suite of ML models. We find that both the Multiple Demand and Language systems--brain systems which are responsible for very different cognitive tasks, encode specific code properties and uniquely align with machine learned representations of code.

Language models have been shown to be very effective in predicting brain recordings of subjects experiencing complex language stimuli.

We study how their representations differ across layer depth, context length, and attention type.

Brain-computer interfaces (BCIs) with speech decoding from brain recordings have broad application potential in fields such as clinical rehabilitation and cognitive neuroscience. However, current decoding methods remain limited to single-language, single-subject, and single neuroimaging modality settings, restricting their clinical applicability and generalizability. Here we propose a joint multilingual, multi-subject and multimodal decoding framework. It maps diverse brain recordings into a unified semantic space defined by a pre-trained multilingual model (PMM), enabling decoding across multiple languages, multiple subjects and multiple neuroimaging modalities. The proposed framework is validated using non-invasive brain recordings from 159 participants across four languages. Experimental results show that it exhibits strong generalization across multilingual, multi-subject, and multimodal settings. More importantly, the proposed framework can promote linguistic fairness, which is vital for underrepresented languages in BCI applications. The unified semantic space enables cross-lingual mapping enhancement, allowing the framework to boost the decoding performance of underrepresented languages, thereby promoting linguistic fairness. Overall, the proposed framework establishes a new potential paradigm for brain decoding, opening new paths for broader applications of BCI.

What aspects of computer programs are represented by the human brain during comprehension? We leverage brain recordings derived from functional magnetic resonance imaging (fMRI) studies of programmers comprehending Python code to evaluate the properties and code-related information encoded in the neural signal. We first evaluate a selection of static and dynamic code properties, such as abstract syntax tree (AST)-related and runtime-related metrics. Then, to learn whether brain representations encode fine-grained information about computer programs, we train a probe to align brain recordings with representations learned by a suite of ML models. We find that both the Multiple Demand and Language systems--brain systems which are responsible for very different cognitive tasks, encode specific code properties and uniquely align with machine learned representations of code.

Recent studies have demonstrated that the representations of artificial neural networks (ANNs) can exhibit notable similarities to cortical representations when subjected to identical auditory sensory inputs. In these studies, the ability to predict cortical representations is probed by regressing from ANN representations to cortical representations. Building upon this concept, our approach reverses the direction of prediction: we utilize ANN representations as a supervisory signal to train recognition models using noisy brain recordings obtained through non-invasive measurements. Specifically, we focus on constructing a recognition model for music identification, where electroencephalography (EEG) brain recordings collected during music listening serve as input. By training an EEG recognition model to predict ANN representations-representations associated with music identification-we observed a substantial improvement in classification accuracy. This study introduces a novel approach to developing recognition models for brain recordings in response to external auditory stimuli. It holds promise for advancing brain-computer interfaces (BCI), neural decoding techniques, and our understanding of music cognition. Furthermore, it provides new insights into the relationship between auditory brain activity and ANN representations.