Research map / Label-efficient and robust learning / Recognition and long-tail
Uncertainty-aware multi-view learning
Fuses views using uncertainty estimates, consensus networks and openness-aware designs.
95 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Recognition and long-tail, keyword cluster “multi view · views · view clustering”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 3 tracked benchmarks.
12 papers in 2023–24 and 51 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 1.0% to 3.9%.
Open questions the data points to
- CUB (ACC): best went from 91.26 (up to 2024) to 95.19 in 2025 — Enhancing Multi-View Classification Reliability with Adaptive Rejection (AAAI 2025 · 13 citations)
- Papers report AE2-Nets on Handwritten (ACC) anywhere from 85.62 to 96.65: comparisons against it depend on whose number you take.
No longer reported by papers here: Caltech101 (ACC, last in 2021).
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 · Multi-View Deep Network for Cross-View Classification (CVPR 2016 · 189 citations)
- 2017 · Multi-view Recurrent Neural Acoustic Word Embeddings (ICLR 2017 · 33 citations)
- 2018 · Latent Discriminant Subspace Representations for Multi-View Outlier Detection (AAAI 2018 · 39 citations)
- 2019 · AE2-Nets: Autoencoder in Autoencoder Networks (CVPR 2019 · 241 citations)
- 2020 · Learning Robust Representations via Multi-View Information Bottleneck (ICLR 2020 · 78 citations)
- 2021 · Multi-View Information-Bottleneck Representation Learning (AAAI 2021 · 93 citations)
- 2022 · Multi-View Intent Disentangle Graph Networks for Bundle Recommendation (AAAI 2022 · 108 citations)
- 2023 · Exploring and Exploiting Uncertainty for Incomplete Multi-View Classification (CVPR 2023 · 43 citations)
- 2024 · DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · 31 citations)
- 2025 · DAMMFND: Domain-Aware Multimodal Multi-view Fake News Detection (AAAI 2025 · 26 citations)
- 2026 · Trusted Multi-view Learning for Long-tailed Classification (AAAI 2026 · 3 citations)
Earliest papers here
- Multi-View Deep Network for Cross-View Classification (CVPR 2016 · 189 citations)
- Robust Multi-View Subspace Learning through Dual Low-Rank Decompositions (AAAI 2016 · 79 citations)
- Information Bottleneck Learning Using Privileged Information for Visual Recognition (CVPR 2016 · 59 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
CUB · ACC (higher is better)
6 papers, 2021–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 91 | Trusted Multi-View Classification (ICLR 2021 · 48 citations) |
| 2023 | 91.26 | Progressive Deep Multi-View Comprehensive Representation Learning (AAAI 2023 · 17 citations) |
| 2025 | 95.19 | Enhancing Multi-View Classification Reliability with Adaptive Rejection (AAAI 2025 · 13 citations) |
Caltech101 · ACC (higher is better)
4 papers, 2020–2021 · not reported since 2021
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 76 | Deep Embedded Complementary and Interactive Information for Multi-View Classification (AAAI 2020 · 62 citations) |
| 2021 | 92.93 | Trusted Multi-View Classification (ICLR 2021 · 48 citations) |
Handwritten · ACC (higher is better)
4 papers, 2021–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 98.51 | Trusted Multi-View Classification (ICLR 2021 · 48 citations) |
| 2023 | 98.97 | Progressive Deep Multi-View Comprehensive Representation Learning (AAAI 2023 · 17 citations) |
What has been tried
62 papers here report 584 measured changes (a component added, removed or swapped, compared with a baseline): helped 79%, no clear effect 5%, hurt 12%, 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 |
|---|---|---|---|---|
| Add view cross-entropy and view contrastive losses | PKU-MMD cross-subject · accuracy | 94.0 -> 95.8 | Helped | DVANet: Disentangling View and Action Features for Multi-View Action Recognition (AAAI 2024 · 31 citations) |
| add Multi-channel Features Aggregation | Weibo · accuracy | 0.934 -> 0.944 | Helped | DAMMFND: Domain-Aware Multimodal Multi-view Fake News Detection (AAAI 2025 · 26 citations) |
| Add local GCN neighborhood feature extraction | HandWritten normal · accuracy (%) | 95.75 -> 99.20 | Helped | Trusted Unified Feature-Neighborhood Dynamics for Multi-View Classification (AAAI 2025 · 21 citations) |
| Add multi-view fusion | ANIMAL, HAND, CUB, SCENE, MRNet, CAL · ACC (%) | 90.46±0.60 -> 94.17±0.20 (ANIMAL); 94.44±0.83 -> 98.23±0.35 (HAND); 87.28±0.46 -> 93.74±1.31 (CUB); 69.08±1.05 -> 73.95± | Helped | Enhancing Multi-View Classification Reliability with Adaptive Rejection (AAAI 2025 · 13 citations) |
| Vary the filtering parameter α from 0.005 to 0.100 | CAL with corrupted views · ACC (%) | 53.75 ± 1.75 -> 53.75 ± 1.75; 55.13 ± 0.38 -> 78.63 ± 1.38 | Mixed | Enhancing Testing-Time Robustness for Trusted Multi-View Classification in the Wild (CVPR 2025 · 12 citations) |
| add cross-view location alignment to the baseline | University-1652 drone to satellite · R@1 | 59.10 -> 84.55 | Helped | From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · 10 citations) |
| Add adaptive imputation, jointly learned with feature selection | WebKB (40% missing, 40% features) · ACC | 55.40 -> 60.85 | Helped | TIME-FS: Joint Learning of Tensorial Incomplete Multi-View Unsupervised Feature Selection and Missing-View Imputation (AAAI 2025 · 10 citations) |
| Add implicit social relations | Yelp, Ciao, and Douban · Recall@5 | Yelp 0.05228 -> 0.05345; Ciao 0.02969 -> 0.03017; Douban 0.08484 -> 0.08503 | Mixed | Robust Graph Based Social Recommendation Through Contrastive Multi-View Learning (AAAI 2025 · 8 citations) |
| add a BEV perspective to FV | nuScenes R1 · mean rotation error (°) | 0.657 -> 0.575 | Helped | RC-AutoCalib: An End-to-End Radar-Camera Automatic Calibration Network (CVPR 2025 · 8 citations) |
| add the DN-Module | ESP-Game · CCR at FPR 10% | 42.56% to 58.65% | Helped | OpenViewer: Openness-Aware Multi-View Learning (AAAI 2025 · 6 citations) |
| add the sufficiency-view constraint (L_SU) | DEFACTO-12 and SSRGFD · F1 | 0.8335 -> 0.9249; 0.6783 -> 0.7995 | Helped | SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints (AAAI 2025 · 6 citations) |
| Add Correlated Multi-view Extractor instead of CGC shared and specific experts | Industrial dataset, 5 clusters · CTR AUC | 0.7692 ± 4.3e-5 -> 0.7698 ± 1.1e-4 | Helped | Multi-Domain Deep Learning from a Multi-View Perspective for Cross-Border E-commerce Search (AAAI 2024 · 6 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| AE2-Nets | Handwritten | ACC | 85.62 – 96.65 | 2 papers, e.g. AAAI 2021 and AAAI 2023 |
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
- Scene15 · 7 of 28 papers
- CUB · 6 of 28 papers
- PIE · 5 of 28 papers
- HandWritten · 4 of 28 papers
- Animal · 3 of 28 papers
- Caltech101 · 3 of 28 papers
- HAND · 2 of 28 papers
- SCENE · 2 of 28 papers
Baselines
- TMC · 7 of 28 papers
- DCCAE · 5 of 28 papers
- ETMC · 5 of 28 papers
- ECML · 5 of 28 papers
- CCML · 4 of 28 papers
- DCCA · 3 of 28 papers
- CPM-Nets · 3 of 28 papers
- DUA-Nets · 3 of 28 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 63 papers; from OpenAlex and the papers' affiliations).
- National University of Defense Technology · 3 papers
- Zhejiang University · 2 papers
- Sichuan University · 2 papers
- Alibaba Group · 2 papers
- Beijing Jiaotong University · 2 papers
- Harbin Institute of Technology · 2 papers
- The Hong Kong Polytechnic University · 2 papers
- Shenzhen Campus of Sun Yat-sen University · 2 papers
- Fuzhou University · 2 papers
- Chinese Academy of Sciences · 2 papers
Use it
Download BibTeX of 40 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
- Incomplete multi-view multi-label learning (35 papers)
Elsewhere in the topic
- Multilingual language representation pretraining (507 papers)
- Contrastive representation theory (495 papers)
- Dense features and foundation priors (482 papers)
- Correspondence and causal representation learning (447 papers)
- Data augmentation and generalization bias (409 papers)
- Partial and multi-label learning (345 papers)
- Loss design and classifier theory (273 papers)
- Group-robust classification (251 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).