Research map / Vision-language and multimodal / Human and embodied modeling
Reducing Semantic Mismatch in Brain-to-Text Decoding Through Personalized Multimodal Masking
Jiaxuan Chen, Yu Qi, Yueming Wang, Gang Pan
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 |
|---|---|---|---|
| add OT-driven semantic masking (ablation) | Yo'Mind · NSD S1 (ventral, lateral, parietal ROIs) | CIDEr: 46.87 -> 58.87 | Improved |
| add OT masking with ventral voxels only (ablation) | Yo'Mind · NSD S1 (ventral ROI) | CIDEr: 43.68 -> 46.24 | Improved |
| add OT masking with ventral+lateral voxels (ablation) | Yo'Mind · NSD S1 (ventral+lateral ROIs) | CIDEr: 45.47 -> 52.80 | Improved |
| add parietal to ventral+lateral voxels with masking (ablation) | Yo'Mind · NSD S1 | CIDEr: 52.80 -> 58.87 | Improved |
| add text semantics to visual semantic set (ablation) | Yo'Mind · NSD S1 | CIDEr: 57.24 -> 58.87 | Improved |
| add visual semantics to text-only set (ablation) | Yo'Mind · NSD S1 | CIDEr: 45.02 -> 58.87 | Improved |
| use Harmon captions instead of SMALLCAP (ablation) | Yo'Mind · NSD S1 | CIDEr: 57.21 -> 58.87 | Improved |
| use BLIP3o captions instead of Harmon (ablation) | Yo'Mind · NSD S1 | CIDEr: 58.87 -> 58.47 | No clear effect |
| use 8 fMRI patches instead of 4 (ablation) | Yo'Mind · NSD (all subjects) | CIDEr: 52.78 -> 56.07 | Improved |
| use 8 fMRI patches instead of 16 (ablation) | Yo'Mind · NSD (all subjects) | CIDEr: 55.14 -> 56.07 | Improved |
| replace fixed hard patch masking (Mind-SA) with OT soft masking | Yo'Mind · NSD brain captioning (COCO captions) | BLEU-4: 18.86 -> 20.88 | Improved |
| replace MindGPT with OT-masked decoder | Yo'Mind · NSD brain captioning (COCO captions) | METEOR: 20.04 -> 38.40 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- UMBRAE — UMBRAE: Unified Multimodal Brain Decoding (ECCV 2024) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- $i$MIND: Insightful Multi-subject Invariant Neural Decoding (NeurIPS 2025 · similarity 0.90)
- Multigranular Evaluation for Brain Visual Decoding (AAAI 2026 · similarity 0.89)
- Breaking the Illusion: When Positive Meets Negative in Multimodal Decoding (CVPR 2026 · similarity 0.88)
- MindAlign: Bridging EEG, Vision, and Language for Zero-Shot Visual Decoding (NeurIPS 2026 · similarity 0.87)
- CLIP-MUSED: CLIP-Guided Multi-Subject Visual Neural Information Semantic Decoding (ICLR 2024 · similarity 0.87)
- Correlating instruction-tuning (in multimodal models) with vision-language processing (in the brain) (ICLR 2025 · similarity 0.87)
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.
