Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-supervised neuroimaging representations
Learns representations of brain signals and imaging with self-supervised objectives.
98 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Self-supervised and contrastive, keyword cluster “depth · self supervision · pre”
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 55 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 7.1% to 17.1%.
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
- 2017 · LSTM Self-Supervision for Detailed Behavior Analysis (CVPR 2017 · 51 citations)
- 2018 · The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (CVPR 2018 · 13,855 citations)
- 2020 · Auto-GAN: Self-Supervised Collaborative Learning for Medical Image Synthesis (AAAI 2020 · 88 citations)
- 2021 · Your head is there to move you around: Goal-driven models of the primate dorsal pathway (NeurIPS 2021 · 54 citations)
- 2022 · AP-BSN: Self-Supervised Denoising for Real-World Images via Asymmetric PD and Blind-Spot Network (CVPR 2022 · 163 citations)
- 2023 · LG-BPN: Local and Global Blind-Patch Network for Self-Supervised Real-World Denoising (CVPR 2023 · 69 citations)
- 2024 · EEGPT: Pretrained Transformer for Universal and Reliable Representation of EEG Signals (NeurIPS 2024 · 35 citations)
- 2025 · AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities (CVPR 2025 · 28 citations)
- 2026 · CalM: A Self-Supervised Foundation Model for Population Dynamics in Calcium Imaging Data (ICML 2026 · 1 citation)
Earliest papers here
- LSTM Self-Supervision for Detailed Behavior Analysis (CVPR 2017 · 51 citations)
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (CVPR 2018 · 13,855 citations)
- Auto-GAN: Self-Supervised Collaborative Learning for Medical Image Synthesis (AAAI 2020 · 88 citations)
What has been tried
61 papers here report 652 measured changes (a component added, removed or swapped, compared with a baseline): helped 70%, no clear effect 9%, hurt 18%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Remove alignment loss | BCIC-2A · Balanced Accuracy | 0.5846±0.0070 -> 0.5287±0.0086 | Hurt | EEGPT: Pretrained Transformer for Universal and Reliable Representation of EEG Signals (NeurIPS 2024 · 35 citations) |
| replace block masking with random token dropping | PASTIS-HD Fold 1 classification · macro F1 | 72.0 -> 71.3 | Hurt | AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalities (CVPR 2025 · 28 citations) |
| use participant-level rather than segment-level positive pairs | Age classification · AUC | 0.900 -> 0.976 | Helped | Large-scale Training of Foundation Models for Wearable Biosignals (ICLR 2024 · 23 citations) |
| add sparse convolutions and mask-aware normalization | Development datasets (D1-D5) · Avg. DSC | 71.35 -> 71.36 | No clear effect | Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · 19 citations) |
| Use all available pre-event images rather than a shorter time series | Kuro Siwo GRD test set · F1-score | Helped | Kuro Siwo: 33 billion $m^2$ under the water. A global multi-temporal satellite dataset for rapid flood mapping (NeurIPS 2024 · 14 citations) | |
| Add pretraining rather than training the MOTOR architecture from scratch | EHR-OMOP code-based tasks · time-dependent C statistic | Celiac: 0.696 -> 0.802; Heart Attack: 0.795 -> 0.887; Lupus: 0.803 -> 0.863; NAFLD: 0.821 -> 0.864; Pancreatic Cancer: 0 | Helped | MOTOR: A Time-to-Event Foundation Model For Structured Medical Records (ICLR 2024 · 10 citations) |
| Add cross-channel constraint to sensitivity map estimation | FastMRI knee PD, 1D Gaussian R=8 · PSNR | 32.43 ± 3.08 -> 32.57 ± 3.24 | Helped | AeSPa : Attention-guided Self-supervised Parallel Imaging for MRI Reconstruction (CVPR 2025 · 6 citations) |
| Use ST-GCN and LSTM encoders with InfoNCE rather than the reported encoder/loss alternatives | Dance vocabulary recognition · accuracy | 60.68 ± 0.58 -> 71.43 ± 0.71 | Helped | DanceMVP: Self-Supervised Learning for Multi-Task Primitive-Based Dance Performance Assessment via Transformer Text Prompting (AAAI 2024 · 6 citations) |
| Train at 224 then fine-tune at 448 | Cityscapes short semantic segmentation (ALL) · mIoU | 64.34 -> 71.81 | Helped | DINO-Foresight: Looking into the Future with DINO (NeurIPS 2025 · 5 citations) |
| Train with all three SSL objectives rather than sVRI only | Downstream classification tasks (average) · AUROC | 0.64 -> 0.67 | Helped | PaPaGei: Open Foundation Models for Optical Physiological Signals (ICLR 2025 · 5 citations) |
| add latent-space self-supervised pre-training rather than no pre-training | Cardiology · AUPRC (%) | 52.26{ ±3.14} -> 53.84{ ±2.24} | Helped | SMART: Towards Pre-trained Missing-Aware Model for Patient Health Status Prediction (NeurIPS 2024 · 5 citations) |
| Add neuron masking only | single-session selected session · co-smoothing | 0.84 -> 1.04 | Helped | Towards a "Universal Translator" for Neural Dynamics at Single-Cell, Single-Spike Resolution (NeurIPS 2024 · 5 citations) |
What reviewers will expect
Datasets and baselines named in the 28 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Sen1Floods11 · 2 of 28 papers
- WESAD · 2 of 28 papers
- TUAB · 2 of 28 papers
Baselines
- DOFA · 2 of 28 papers
- CROMA · 2 of 28 papers
- SatMAE · 2 of 28 papers
- SPaRCNet · 2 of 28 papers
- MLP · 2 of 28 papers
- POYO+ · 2 of 28 papers
- BYOL · 2 of 28 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 78 papers; from OpenAlex and the papers' affiliations).
- Mila · 5 papers
- McGill University · 4 papers
- Harbin Institute of Technology · 4 papers
- Georgia Institute of Technology · 3 papers
- East China Normal University · 2 papers
- Shanghai Jiao Tong University · 2 papers
- ShanghaiTech University · 2 papers
- Peking University · 2 papers
- University of Wisconsin - Madison · 2 papers
- Stanford University · 2 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
- Pretext-task visual pre-training (136 papers)
- Self-supervision in supervised pipelines (121 papers)
- Self-supervised speech and brain decoding (56 papers)
Elsewhere in the topic
- Invariance-based self-supervised learning (168 papers)
- Data-efficient and robust self-supervision (122 papers)
- Contrastive versus non-contrastive theory (74 papers)
- Inductive bias and minimal representations (66 papers)
- Masked autoencoder pretraining (48 papers)
- Empirical contrastive visual learning (38 papers)
- Representation evaluation and trade-offs (37 papers)
- Self-supervised speech representations (24 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).