Research map / 3D vision / Face and shape
Beyond Grid-Locked Voxels: Neural Response Functions for Continuous Brain Encoding
Haomiao Chen, Keith W Jamison, Mert R. Sabuncu, Amy Kuceyeski
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune a pretrained NRF and voxel-wise ensemble instead of training NRF from scratch (ablation) | NRF, new subject 7 · 200 images | voxel Pearson r: 0.180 -> 0.227 | Improved |
| fine-tune a pretrained NRF with ensemble instead of scratch (ablation) | NRF, new subject 7 · 20 images | voxel MSE: 0.417 -> 0.445 | Hurt |
| use voxel-wise regression ensemble instead of simple averaging of fine-tuned models (ablation) | NRF finetune ensemble · new subject 7, 200 images | CLIP two-way identification: 74.9% -> 87.3% | Improved |
| use simple averaging instead of voxel-wise regression ensemble (ablation) | NRF finetune ensemble · new subject 7, 200 images | voxel Pearson r: 0.227 -> 0.253 | Improved |
| ensemble three fine-tuned base models instead of a single fine-tuned base model (ablation) | NRF finetune · new subject 7, 200 images | PixCorr: 0.226-0.246 -> 0.255 | Improved |
| shuffle coordinate-response pairings (global or within ROI) to break local smoothness (ablation) | NRF single subject · low-data regimes | voxel Pearson r | Hurt |
| shift MNI coordinates between subjects during finetuning to break alignment (ablation) | NRF cross-subject transfer · subject 1 to subject 7, EBA and V1 | voxel Pearson r | Hurt |
| condition predictions on MNI coordinates (NRF) instead of flattened voxel vector outputs | NRF vs linear and fWRF · low-data single-subject training | voxel Pearson r | Improved |
| use NRF instead of fWRF on semantic-level metrics via MindEye2 | NRF · full data | semantic-level (PixCorr, CLIP): PixCorr 0.303 (fWRF) vs 0.261; CLIP 91.9% vs 89.3% | No clear effect |
Closest papers by meaning
- Brain Harmony: A Multimodal Foundation Model Unifying Morphology and Function into 1D Tokens (NeurIPS 2025 · similarity 0.87)
- FlatClip: Reusing Image Foundation Models for fMRI Representation Learning via Cortical Flatmaps (NeurIPS 2026 · similarity 0.87)
- Meta-Learning an In-Context Transformer Model of Human Higher Visual Cortex (NeurIPS 2025 · similarity 0.87)
- MIRAGE: Adaptive Multimodal Gating for Whole-Brain fMRI Encoding (NeurIPS 2026 · similarity 0.87)
- BrainLM: A foundation model for brain activity recordings (ICLR 2024 · similarity 0.87)
- MindBridge: A Cross-Subject Brain Decoding Framework (CVPR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
