Research map / Vision-language and multimodal / Human and embodied modeling
fMRI visual decoding
Decodes and synthesizes visual stimuli from brain recordings using semantic and cross-subject models.
106 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Human and embodied modeling, keyword cluster “concepts · brain · fmri”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
23 papers in 2023–24 and 68 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 4.3% to 10.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · How Deep is the Feature Analysis underlying Rapid Visual Categorization? (NeurIPS 2016 · 29 citations)
- 2019 · A Neurobiological Evaluation Metric for Neural Network Model Search (CVPR 2019 · 15 citations)
- 2020 · High-contrast “gaudy” images improve the training of deep neural network models of visual cortex (NeurIPS 2020 · 4 citations)
- 2021 · Disentangling syntax and semantics in the brain with deep networks (ICML 2021 · 32 citations)
- 2022 · Mind Reader: Reconstructing complex images from brain activities (NeurIPS 2022 · 35 citations)
- 2023 · Brain encoding models based on multimodal transformers can transfer across language and vision (NeurIPS 2023 · 16 citations)
- 2024 · Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction (NeurIPS 2024 · 9 citations)
- 2025 · In vivo cell-type and brain region classification via multimodal contrastive learning (ICLR 2025 · 11 citations)
- 2026 · HyFI: Hyperbolic Feature Interpolation for Brain-Vision Alignment (AAAI 2026 · 1 citation)
Earliest papers here
- How Deep is the Feature Analysis underlying Rapid Visual Categorization? (NeurIPS 2016 · 29 citations)
- “Congruent” and “Opposite” Neurons: Sisters for Multisensory Integration and Segregation (NeurIPS 2016 · 9 citations)
- A Neurobiological Evaluation Metric for Neural Network Model Search (CVPR 2019 · 15 citations)
What has been tried
59 papers here report 590 measured changes (a component added, removed or swapped, compared with a baseline): helped 63%, no clear effect 9%, hurt 20%, mixed 7%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add blur prior that reduces high-frequency image detail | THINGS-EEG · top-1 accuracy | 42.1 -> 49.3 | Helped | Bridging the Vision-Brain Gap with an Uncertainty-Aware Blur Prior (CVPR 2025 · 10 citations) |
| Use ViT3D rather than flattening and patching fMRI data | NSD brain caption · BLEU4 | 8.49 -> 15.85 | Helped | Neuro-Vision to Language: Enhancing Brain Recording-based Visual Reconstruction and Language Interaction (NeurIPS 2024 · 9 citations) |
| Use leave-one-out CL instead of pairwise CL | Sleep stage classification · AUROC | 0.876 -> 0.906 | Helped | SleepFM: Multi-modal Representation Learning for Sleep Across Brain Activity, ECG and Respiratory Signals (ICML 2024 · 6 citations) |
| Add LoRAs to the fine-tuned model | NSD, Subject 1, 1-hour fine-tuning · PixCorr | 0.235 -> 0.261 | Helped | MindTuner: Cross-Subject Visual Decoding with Visual Fingerprint and Semantic Correction (AAAI 2025 · 4 citations) |
| Train the decoder with cross-subject fMRI data rather than single-subject data | NSD subject 1 · PixCorr | .417 -> .445 | Helped | See Through Their Minds: Learning Transferable Brain Decoding Models from Cross-Subject fMRI (AAAI 2025 · 4 citations) |
| Add MoBE adapters and corresponding meta-learning | NSD multi-label classification · mAP | 0.361 -> 0.424 | Helped | Wills Aligner: Multi-Subject Collaborative Brain Visual Decoding (AAAI 2025 · 3 citations) |
| Add Past State features to the random-feature baseline | Visnav · R² | 0.0521 -> 0.2299 | Helped | Neuroformer: Multimodal and Multitask Generative Pretraining for Brain Data (ICLR 2024 · 3 citations) |
| Add core techniques, tune hyperparameters, and train on ten mice instead of five | Main track live-test · score | ≈0.19 -> 0.25 | Helped | Retrospective for the Dynamic Sensorium Competition for predicting large-scale mouse primary visual cortex activity from videos (NeurIPS 2024 · 3 citations) |
| Add the orthogonal constraint to low- and high-level RSA guidance | NSD · mAP | .201±.006 -> .258±.017 | Helped | CLIP-MUSED: CLIP-Guided Multi-Subject Visual Neural Information Semantic Decoding (ICLR 2024 · 2 citations) |
| Add within-participant cluster expansion | NSD · cluster location and semantic interpretation | No clear effect | Finding Shared Decodable Concepts and their Negations in the Brain (ICLR 2025 · 1 citation) | |
| Add signal reconstruction loss L_rec | NSD, Subject 2 novel / Subject 1 known · PixCorr | 0.072 -> 0.186 | Helped | MindAligner: Explicit Brain Functional Alignment for Cross-Subject Visual Decoding from Limited fMRI Data (ICML 2025 · 0 citations) |
| add Brain Instruction Tuning | TDIUC task adaptation · H-MPT | 19.38% -> 45.43% | Helped | MindLLM: A Subject-Agnostic and Versatile Model for fMRI-to-text Decoding (ICML 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 35 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Natural Scenes Dataset (NSD) · 14 of 35 papers
- NSD · 4 of 35 papers
- COCO · 4 of 35 papers
- MSCOCO · 3 of 35 papers
- THINGS-MEG · 3 of 35 papers
- ImageNet · 3 of 35 papers
- THINGS-EEG · 2 of 35 papers
- Natural Scenes Dataset · 2 of 35 papers
Baselines
- MindEye2 · 8 of 35 papers
- MindBridge · 6 of 35 papers
- Brain Diffuser · 4 of 35 papers
- MindEye1 · 3 of 35 papers
- UMBRAE · 3 of 35 papers
- BraVL · 3 of 35 papers
- CLIP-MUSED · 3 of 35 papers
- NICE · 3 of 35 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 91 papers; from OpenAlex and the papers' affiliations).
- Chinese Academy of Sciences · 6 papers
- Massachusetts Institute of Technology · 4 papers
- Columbia University · 4 papers
- Harvard University · 4 papers
- The University of Hong Kong · 4 papers
- Carnegie Mellon University · 3 papers
- Zhejiang University · 3 papers
- Tsinghua University · 3 papers
- The Chinese University of Hong Kong · 3 papers
- Chinese academy of science · 3 papers
Use it
Download BibTeX of 36 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Language-grounded visual concept learning (83 papers)
- Concept bottleneck models (21 papers)
Elsewhere in the topic
- Vision-language pretraining (182 papers)
- Medical vision-language models (130 papers)
- Scene graph generation (126 papers)
- Language-grounded scene text and structure (124 papers)
- Human-like visual cognition in VLMs (97 papers)
- Multimodal and affective interaction modeling (87 papers)
- Multimodal retrieval-augmented modeling (85 papers)
- Composed and text-based image retrieval (78 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).