Search:
Match:
1 results

SLIM-Brain: Efficient fMRI Foundation Model

Published:Dec 26, 2025 06:10
1 min read
ArXiv

Analysis

This paper introduces SLIM-Brain, a novel foundation model for fMRI analysis designed to address the data and training inefficiency challenges of existing methods. It achieves state-of-the-art performance on various benchmarks while significantly reducing computational requirements and memory usage compared to traditional voxel-level approaches. The two-stage adaptive design, incorporating a temporal extractor and a 4D hierarchical encoder, is key to its efficiency.
Reference

SLIM-Brain establishes new state-of-the-art performance on diverse tasks, while requiring only 4 thousand pre-training sessions and approximately 30% of GPU memory comparing to traditional voxel-level methods.