Research map / Vision-language and multimodal / Human and embodied modeling
Linguistic Priors for Visual Decoupling: Towards Symmetric Vision-Brain Alignment
Dongjun Liu, Weichen Dai, Jingsheng Qian, Honggang Liu, Hangjie Yi, Wanzeng Kong
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 text-guided visual decoupling (non-uniform center-sharp Gaussian blur) to vanilla contrastive alignment (ablation) | EEG-image contrastive alignment · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 43.6 -> 54.9 | Improved |
| add text-guided visual decoupling to vanilla contrastive alignment (ablation) | EEG-image contrastive alignment · THINGS-EEG (intra-subject) | Top-5 accuracy (%): 77.0 -> 87.7 | Improved |
| add dynamic alignment strategy on top of decoupling (ablation) | Decouple configuration · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 54.9 -> 55.6 | Improved |
| add dynamic alignment strategy on top of decoupling (ablation) | Decouple configuration · THINGS-EEG (intra-subject) | Top-5 accuracy (%): 87.7 -> 87.5 | No clear effect |
| add multimodal similarity modulation (semantic enhancement) (ablation) | Dynamic configuration · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 55.6 -> 58.2 | Improved |
| add multimodal similarity modulation (semantic enhancement) (ablation) | Dynamic configuration · THINGS-EEG (intra-subject) | Top-5 accuracy (%): 87.5 -> 89.1 | Improved |
| increase central blur level c from 0.1 to 0.5 (ablation) | Ours · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 56.2 -> 58.2 | Improved |
| vary c between 0.2 and 0.5 (ablation) | Ours · THINGS-EEG (intra-subject) | Top-5 accuracy (%): 88.1 to 89.1 | No clear effect |
| swap CLIP RN50 image encoder for RN101 (ablation) | Ours · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 58.2 -> 55.8 | Hurt |
| swap CLIP RN50 image encoder for ViT-B/16 (ablation) | Ours · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 58.2 -> 54.5 | Hurt |
| swap CLIP RN50 image encoder for ViT-B/32 (ablation) | Ours · THINGS-EEG (intra-subject) | Top-1 accuracy (%): 58.2 -> 58.0 | No clear effect |
| replace UBP with the proposed method | THINGS-EEG inter-subject retrieval · THINGS-EEG (inter-subject, leave-one-out) | Top-1 accuracy (%): 12.4 -> 15.8 | Improved |
Closest papers by meaning
- MindAlign: Bridging EEG, Vision, and Language for Zero-Shot Visual Decoding (NeurIPS 2026 · similarity 0.90)
- Multigranular Evaluation for Brain Visual Decoding (AAAI 2026 · similarity 0.90)
- Neural‑Visual Decoding via Cognitive‑guided Adaptive Blurring and Information‑Constrained Alignment (NeurIPS 2026 · similarity 0.90)
- Human-Aligned Image Models Improve Visual Decoding from the Brain (ICML 2025 · similarity 0.89)
- Brain decoding: toward real-time reconstruction of visual perception (ICLR 2024 · similarity 0.89)
- Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction (NeurIPS 2024 · similarity 0.89)
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.
