리서치 레이더

이번 주 우리가 읽고 있는 것

뇌 파운데이션 모델, fMRI/EEG 역학, 유전체학과 커넥토믹스, 양자 머신러닝 분야의 새 arXiv 논문을 매주 스캔합니다. 요약은 LLM 파이프라인이 생성하며 생성 방식이 함께 표기됩니다. 연구실 논문도 선별 수록됩니다.

이 논문들이 우리에게 어떤 의미일지 궁금하다면, AI 아이디어 랩 을 방문해 보세요 — 이 레이더를 바탕으로 매주 기계가 제안하는 가설들입니다.

12건 · arXiv에서 매주 갱신

EEG 파운데이션 모델Emerging Paper

2026-08-13

EEG-PRIME: Prototype-Aligned Representation Learning with Multi-Level Conditioning for EEG Decoding

arXiv (2026-08) · Shuailei Zhang, Muyun Jiang, Wei Zhang, Jinbo Chen

요약

•EEG-PRIME is a novel two-stage foundation model for EEG decoding designed to overcome cross-dataset and cross-subject generalization issues.
•It employs masked pretraining with spectral augmentation followed by prototype-aligned instruction tuning, incorporating task-semantic, dataset-specific, and subject-invariant conditioning for robust decoding.
•EEG-PRIME demonstrates superior performance on diverse EEG tasks and exhibits strong zero-shot transfer capabilities on unseen datasets, outperforming existing baselines and foundation models.

의의: This work represents a significant advancement in EEG decoding by developing a foundation model capable of robust and generalizable performance across diverse datasets and subjects without extensive calibration, paving the way for more practical brain-computer interfaces.

우리 연구실과의 연결: The development of EEG-PRIME aligns with the SNU Connectome Lab's Neuro-X initiative, as it focuses on creating generalizable AI models for neural signal decoding, a critical component for advancing neurotechnology.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-13

EEG-PRISM: Physiologically-Grounded Interpretability of Predictions by EEG Foundation Models

arXiv (2026-08) · Deeksha M Shama, Punnisa Amornsirikul, Archana Venkataraman

요약

•EEG-PRISM is a novel method that translates time-channel attribution scores from EEG foundation models into more physiologically relevant domains, specifically the frequency and source domains, without altering the original model.
•The method utilizes linear transformations, including an invertible DFT for frequency mapping and an approximately invertible EEG generative model for source localization, to provide clinically intuitive interpretations.
•Evaluations in both simulated and real EEG data (epilepsy and autism) demonstrated EEG-PRISM's effectiveness in recovering ground-truth spectral activity and localizing seizure onset regions and predictive biomarkers with reasonable accuracy.

의의: This work introduces a crucial tool for making advanced EEG foundation models interpretable in a clinically meaningful way, thereby enabling better understanding of neural mechanisms and identification of disease biomarkers.

우리 연구실과의 연결: EEG-PRISM's focus on interpretable AI for EEG foundation models directly aligns with SNU Connectome Lab's Neuro-X initiative, which aims to develop advanced neurotechnology with a strong emphasis on understanding and explaining neural data.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-12

Continuous-Latent Predictive Modeling with Semantic Alignment for EEG-Language Foundation Models

arXiv (2026-08) · Myeong-Ju Cho, Hye-Bin Shin, Seo-Hyun Lee, Seong-Whan Lee

요약

•The paper proposes Brain Latent Predictive Model (BLPM), an EEG-language foundation model that redefines EEG decoding as a continuous semantic embedding prediction problem.
•BLPM introduces a Continuous EEG Latent Predictive (CELP) encoder for learning transferable representations and a Multi-Query Semantic Decomposition (MQSD) module for aligning EEG with textual semantics in a shared latent space.
•The model aims to overcome limitations of masked autoencoding and autoregressive modeling in current EEG foundation models by prioritizing task-relevant semantics and addressing the mismatch between continuous neural dynamics and discrete token spaces.

의의: This work establishes a novel paradigm for EEG-language foundation models, demonstrating improved generalization and alignment of continuous EEG signals with natural language semantics.

우리 연구실과의 연결: This research is highly relevant to the 'Large Brain Models' and 'Neuro-X' initiatives at the SNU Connectome Lab, as it advances the development of foundation models for integrating brain activity with artificial intelligence and language processing.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-07

ZIPBrain: Can EEG Foundation Models Be Faster, Locally Deployable, but Accurate?

arXiv (2026-08) · Lingwei Li, Yirong Kan, Peng Chen, Xu Cao

요약

•ZIPBrain is introduced as a novel redundancy-aware EEG token pooling module designed to make EEG Foundation Models (EFMs) faster and more deployable.
•It addresses the computational burden of EFMs by reducing the token count through a partitioning and merging strategy, leveraging EEG's low Signal-to-Noise Ratio (SNR) characteristics.
•ZIPBrain is a plug-and-play module that improves EFM accuracy (1.3%-10.5%) and significantly reduces inference time (32.7% average, up to 41.8% with CUDA Graph) without substantial computational overhead.

의의: This work demonstrates a crucial advancement in making powerful EEG foundation models practical for real-time, resource-constrained clinical applications by efficiently managing computational demands while enhancing performance.

우리 연구실과의 연결: ZIPBrain's focus on accelerating and deploying large EEG models without sacrificing accuracy directly aligns with the SNU Connectome Lab's research into 'Large Brain Models' and could be integrated into our 'Neuro-X' initiatives for more efficient neurological data processing.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-05

SW-ProxyCE: Zero-Query Adversarial Transfer from Public EEG Encoders to Private Downstream Models

arXiv (2026-08) · Linhua Cong, Dingkun Liu, Dongrui Wu

요약

•This paper introduces a novel security risk in EEG foundation models: adversarial transfer attacks from public encoders to private downstream models.
•They propose SW-ProxyCE, a query-free attack framework that leverages shrinkage-whitened class prototypes from a small labeled reference set to generate transferable adversarial examples.
•SW-ProxyCE effectively attacks inaccessible downstream models across various EEG tasks and foundation encoders, outperforming task-agnostic attacks and demonstrating that foundation model transferability does not imply adversarial robustness.

의의: This research highlights a critical security vulnerability in the deployment of EEG foundation models, urging the development of robust defenses to protect private downstream applications.

우리 연구실과의 연결: This work is highly relevant to Neuro-X, as it addresses critical security and robustness challenges pertinent to the deployment and trustworthiness of large-scale neural data models.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-03

STEAM: A Spatio-TEmporal Alignment Mixture-of-Experts Model with Hierarchical Pre-training for EEG Decoding

arXiv (2026-08) · Zhu Chen, Dingkun Liu, Yuheng Chen, Dongrui Wu

요약

•STEAM is a novel spatio-temporal alignment Mixture-of-Experts (MoE) model designed for EEG decoding, addressing challenges in generalizability and adaptation costs of conventional BCI algorithms.
•It employs a dual-branch spatio-temporal encoder with a Shared Soft Mixture-of-Experts (SSMoE) module to align and exchange information between spatial and temporal representations.
•The model utilizes a hierarchical pre-training strategy that allows for general initialization and subsequent specialization to specific EEG paradigms, achieving superior performance across diverse datasets and evaluation settings.

의의: This work introduces a foundational model for EEG decoding that significantly advances the generalizability and adaptation efficiency of BCI systems, crucial for widespread adoption in clinical and rehabilitative applications.

우리 연구실과의 연결: STEAM aligns with the Connectome Lab's focus on Large Brain Models and Neuro-X by developing a sophisticated deep learning architecture for robust and generalizable neural signal decoding, applicable to understanding and leveraging complex brain activity.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-08-03

Understanding and Correcting Low-Frequency Bias in EEG Foundation Model

arXiv (2026-08) · Junjie Yu, Zihan Deng, Jianyu Zhang, Junrong Mu

요약

•EEG foundation models exhibit a pervasive low-frequency bias in their learned representations, irrespective of data scale, model capacity, or pretraining objective.
•This bias stems from the interplay between EEG's $1/f^α$ spectral characteristics and neural networks' natural inclination to prioritize low-frequency components, exacerbated by $\ell_2$ reconstruction losses in masked autoencoders.
•FAME (Frequency-balanced Masked Autoencoding) is proposed, which balances spectral supervision by reconstructing time-frequency activity in predefined EEG bands with independent standardization and equal loss weighting, leading to improved, spectrally balanced representations and state-of-the-art performance on 24 out of 41 downstream tasks.

의의: This research identifies a critical and persistent spectral bias in EEG foundation models, offering a novel frequency-balanced solution that significantly improves representation learning and downstream task performance, pushing the boundaries of transferable EEG AI.

우리 연구실과의 연결: This work directly impacts the development of Large Brain Models and Neuro-X initiatives at SNU Connectome Lab by providing crucial insights and a solution for training more effective and balanced EEG foundation models, which are fundamental components for understanding brain activity across diverse tasks and populations.

#EEG

EEG 파운데이션 모델Emerging Paper

2026-07-31

EEG-JEPA: Structured Latent Prediction for EEG Foundation Models

arXiv (2026-07) · Jinhao Li, Zhiyuan Ma, Xueqiao Han, Zhongye Xia

요약

•EEG-JEPA introduces a structured latent-prediction framework for EEG foundation modeling, moving beyond simple masked waveform reconstruction to infer latent states.
•The framework employs a masked context encoder and predictor to infer contextual latent states from a target encoder observing complete input, utilizing Neurotopology-Aware Multi-scale Electrode-Temporal Masking (N-MET) for structured electrode-time prediction.
•EEG-JEPA demonstrates significant improvements in frozen multitask transfer (from 40.49% to 50.42%) and full fine-tuning (from 68.98% to 70.65%) compared to masked waveform reconstruction, setting new benchmarks on EEG-FM-Bench.

의의: This work establishes a more effective pretraining strategy for EEG foundation models, enabling them to learn more transferable and neurally relevant representations by predicting structured latent states rather than noisy raw signals.

우리 연구실과의 연결: This research is highly relevant to the 'Large Brain Models' focus of the SNU Connectome Lab, as it develops advanced foundation models for EEG, a key modality for understanding brain function, and optimizes their pretraining for better generalizability and transferability across various tasks.

#EEG

fMRI 파운데이션 모델Lab SOTA

2025-12-01

NeuroMamba: A State-Space Foundation Model for Functional MRI

NeurIPS 2025 Workshop (Spotlight) · Jubin Choi et al. (SNU Connectome Lab)

요약

•Replaces quadratic Transformer self-attention with linear-time selective state-space layers (Mamba) for 4D fMRI.
•Efficiently models full-length brain scans across thousands of time steps without aggressive temporal downsampling.
•Outperforms standard 3D CNNs and ViT baselines across multi-site cognitive decoding benchmarks.

의의: Pioneers the first scalable State-Space architecture directly tailored to continuous 4D physiological brain dynamics, establishing a new SOTA for fMRI foundation models.

우리 연구실과의 연결: Led by Connectome Lab PhD student Jubin Choi and Prof. Jiook Cha as part of the core Neuro-X initiative.

#fMRI#4D Spatiotemporal#Mamba

유전자와 뇌Nature Comms

2025-09-26

Polygenic architecture of brain structure and function, behaviors, and psychopathologies in children

Nature Communications (2025) · Yoonjung Yoonie Joo, Bo-Gyeom Kim, Gakyung Kim, Eunji Lee, Jungwoo Seo, Jiook Cha

요약

•Maps genome-wide polygenic influences onto brain structure and function in the ABCD developmental cohort.
•Identifies gene-brain convergence patterns linking polygenic scores to cortical and subcortical phenotypes.
•Characterizes pathways from polygenic risk through brain phenotypes to behaviors and psychopathologies.

의의: Demonstrates population-scale genomic-connectome convergence as a foundation for computational psychiatry.

우리 연구실과의 연결: Flagship Connectome Lab publication in Nature Communications.

#Genomics#fMRI#Cortical Morphology

EEG 파운데이션 모델Lab SOTA

2025-07-01

DIVER-0: Fully Channel-Equivariant EEG Foundation Model

ICML 2025 GenBio Workshop (Spotlight) · Dongyeop Han, Ahhyun Lee, Taeyang Lee, Sebin Lee, Jiook Cha

요약

•Formulates EEG sensor arrays as spatial point clouds with continuous coordinate embeddings.
•Guarantees SE(3)/permutation equivariance across arbitrary electrode channel layouts (8 to 256 channels).
•Enables seamless zero-shot cross-dataset transfer across heterogeneous clinical and research setups.

의의: Solves one of the largest obstacles in bio-signal AI: disparate electrode montages across worldwide hospital & research databases.

우리 연구실과의 연결: Developed by Connectome Lab members at SNU and MILA.

#EEG#Equivariant GNN#Zero-Shot

뇌-언어모델 정렬Lab Preprint

2025-02-18

Mind the Gap: Aligning the Brain with Language Models Requires a Nonlinear and Multimodal Approach

arXiv:2502.12771 · Dongyeop Han, Jiook Cha

요약

•Critiques the traditional linear regression assumption between LLM activations and fMRI responses.
•Shows that nonlinear multimodal mappings recover brain-language correspondence that linear encodings underestimate.
•Connects predictive-processing views of cortex with autoregressive Transformer internal states.

의의: Re-evaluates the mathematical foundation of how artificial neural representations correspond to biological neural activity.

우리 연구실과의 연결: Connectome Lab preprint; theoretical grounding for the Neuro-X Large Brain Model initiative.

#fMRI#LLMs#Representational Geometry