NeurIPS: Neuro-anatomical Inductive Priors for Sphere-based Brain Decoding
The paper presents NeurIPS, a framework designed to enhance surface-based brain decoding by utilizing neuro-anatomical inductive priors. It introduces a Selective ROI Spherical Tokenizer and a Structure-Guided Mixture of Experts to improve efficiency and performance. The framework achieves state-of-the-art results on the Natural Scenes Dataset while ensuring rapid adaptation to new subjects.
- ▪NeurIPS improves surface-based decoding by reframing anatomical variation as a predictive signal.
- ▪The framework establishes a new state-of-the-art for surface decoders with performance comparable to strong 1D baselines.
- ▪NeurIPS converges dramatically faster than previous models, requiring only 20% of the data for new subjects.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.24993 (cs) [Submitted on 24 May 2026] Title:NeurIPS: Neuro-anatomical Inductive Priors for Sphere-based Brain Decoding Authors:Sijin Yu, Zijiao Chen, Zhenyu Yang, Zihao Tan, Jiakun Xu, Zhongliang Liu, Shengxian Chen, Wenxuan Wu, Xiangmin Xu, Xin Zhang View a PDF of the paper titled NeurIPS: Neuro-anatomical Inductive Priors for Sphere-based Brain Decoding, by Sijin Yu and 9 other authors View PDF HTML (experimental) Abstract:Current fMRI decoders face a performance-fidelity trade-off where efficient ID encoders outperform geometrically faithful surface-based models. We argue this is partly driven by inefficient surface tokenization and the failure to use anatomy as a predictive signal.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
