# Lost in Visual Translation: A VLM-Assisted Perceptual-Semantic Coherence Framework for EEG-to-Image Reconstruction

arXiv: https://arxiv.org/abs/2607.12364 · HTML: https://mlsanity.com/paper/2607.12364/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 33 references that are in the map.
- First posted on arXiv 14 Jul 2026.

## Builds on (most cited first, 23 of 23)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Demystifying MMD GANs](https://mlsanity.com/paper/1801.01401/) (2018)
- [Exploring CLIP for Assessing the Look and Feel of Images](https://mlsanity.com/paper/2207.12396/) (2022)
- [Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2305.01569/) (2023)
- [GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment](https://mlsanity.com/paper/2310.11513/) (2023)
- [Image Quality Assessment: Unifying Structure and Texture Similarity](https://mlsanity.com/paper/2004.07728/) (2020)
- [Evaluating Text-to-Visual Generation with Image-to-Text Generation](https://mlsanity.com/paper/2404.01291/) (2024)
- [TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering](https://mlsanity.com/paper/2303.11897/) (2023)
- [VIEScore: Towards Explainable Metrics for Conditional Image Synthesis Evaluation](https://mlsanity.com/paper/2312.14867/) (2023)
- [PieAPP: Perceptual Image-Error Assessment through Pairwise Preference](https://mlsanity.com/paper/1806.02067/) (2018)
- [Reconstructing the Mind's Eye: fMRI-to-Image with Contrastive Learning and Diffusion Priors](https://mlsanity.com/paper/2305.18274/) (2023)
- [Seeing Beyond the Brain: Conditional Diffusion Model with Sparse Masked Modeling for Vision Decoding](https://mlsanity.com/paper/2211.06956/) (2022)
- [Generic decoding of seen and imagined objects using hierarchical visual features](https://mlsanity.com/paper/1510.06479/) (2015)
- [MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data](https://mlsanity.com/paper/2403.11207/) (2024)
- [Natural scene reconstruction from fMRI signals using generative latent diffusion](https://mlsanity.com/paper/2303.05334/) (2023)
- [Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion](https://mlsanity.com/paper/2403.07721/) (2024)
- [EEG2IMAGE: Image Reconstruction from EEG Brain Signals](https://mlsanity.com/paper/2302.10121/) (2023)
- [BrainVis: Exploring the Bridge between Brain and Visual Signals via Image Reconstruction](https://mlsanity.com/paper/2312.14871/) (2023)
- [Natural Image Reconstruction from fMRI using Deep Learning: A Survey](https://mlsanity.com/paper/2110.09006/) (2021)
- [BrainBits: How Much of the Brain are Generative Reconstruction Methods Using?](https://mlsanity.com/paper/2411.02783/) (2024)
- [Brain-to-Image Retrieval and Reconstruction via Multimodal EEG Alignment](https://mlsanity.com/paper/2605.23996/) (2026)

## Abstract (arXiv)

EEG-to-image evaluation should distinguish visual fidelity from recoverable meaning. Yet EEG-derived reconstructions are blurry, distorted, and low-detail, causing SSIM, LPIPS, and CLIP to penalize semantically recoverable outputs or reward plausible but incorrect ones. We analyze 6,855 ground-truth/reconstruction pairs from ATM, ENIGMA, BrainVis, and DreamDiffusion using semantic probes, caption harshness and blind-spot rates, and controlled degradations. Pixel metrics show near-zero correlation with semantic consistency, while representation metrics conflate perceptual and semantic errors. We therefore introduce a BCI-aware framework in which four VLMs assess image pairs through structured questions, producing Tolerant Perceptual Alignment Scores (T-PAS) and Tolerant Semantic Alignment Scores (T-SAS). Their consensus is distilled into the BCI-Coherence Score (BCS), a compact evaluator achieving a T-PAS MAE of 0.079 (r = 0.700) and a T-SAS MAE of 0.082 (r = 0.850) on our data. Human validation shows highly reliable joint coherence judgments, with Cohen's kappa = 0.882 +/- 0.174 and Krippendorff's alpha = 0.882, supporting perceptual-semantic recoverability over generic visual similarity. Code and resources are available at https://sukt03.github.io/BCS/.