NeurIPS 2021 accepted papers by research area
2,508 papers were accepted at NeurIPS 2021 (+607 vs 2020). Here is how they split across 20 research areas, and which topics gained the most ground.
See NeurIPS 2021 in the interactive map
Submitting to NeurIPS? Checklist →
Research areas
| Area | Papers | Share | Change in share vs 2020 |
|---|---|---|---|
| Learning theory and algorithms | 367 | 14.6% | -4.5 pts |
| Probabilistic and scientific modeling | 304 | 12.1% | -2.1 pts |
| Label-efficient and robust learning | 262 | 10.4% | +1.3 pts |
| Architectures and efficiency | 225 | 9.0% | +1.4 pts |
| Reinforcement learning | 216 | 8.6% | steady |
| Optimization | 175 | 7.0% | -1.7 pts |
| Bandits and online learning | 139 | 5.5% | +0.4 pts |
| Graph learning | 129 | 5.1% | +0.2 pts |
| Adversarial robustness and security | 91 | 3.6% | -0.2 pts |
| Multi-agent systems and games | 81 | 3.2% | +0.6 pts |
| Federated learning and privacy | 72 | 2.9% | +0.8 pts |
| Image generation | 67 | 2.7% | steady |
| Large language models | 66 | 2.6% | +1.5 pts |
| Detection and segmentation | 64 | 2.6% | -0.4 pts |
| Vision-language and multimodal | 60 | 2.4% | +1.0 pts |
| 3D vision | 59 | 2.4% | +0.3 pts |
| Video | 49 | 2.0% | +1.1 pts |
| Time series | 30 | 1.2% | +0.2 pts |
| Continual learning | 29 | 1.2% | +0.1 pts |
| Point clouds | 23 | 0.9% | steady |
Fastest-growing topics at NeurIPS 2021
Change in share of NeurIPS papers from 2020 to 2021, topics with at least 10 papers.
- Attention and transformers (Architectures and efficiency): 20 → 58 papers
- Language, safety and interpretability (Large language models): 8 → 32 papers
- Visual and representation RL (Reinforcement learning): 52 → 87 papers
- Video segmentation and tracking (Video): 8 → 26 papers
- Offline RL (Reinforcement learning): 7 → 24 papers
- Human-AI and goal-conditioned (Multi-agent systems and games): 21 → 42 papers
- Self-supervised and contrastive (Label-efficient and robust learning): 17 → 36 papers
- Contextual bandits and Thompson sampling (Bandits and online learning): 17 → 35 papers
- Human pose estimation (3D vision): 3 → 16 papers
- Visual question answering (Vision-language and multimodal): 3 → 14 papers
More venues
Same year: ICML 2021 · ICLR 2021 · CVPR 2021 · AAAI 2021 · ICCV 2021 · TPAMI 2021 · IJCV 2021 · JMLR 2021
NeurIPS in other years: 2016 · 2017 · 2018 · 2019 · 2020 · 2022 · 2023 · 2024 · 2025 · 2026
