ICLR 2023 accepted papers by research area
1,575 papers were accepted at ICLR 2023 (+480 vs 2022). Here is how they split across 20 research areas, and which topics gained the most ground.
See ICLR 2023 in the interactive map
Research areas
| Area | Papers | Share | Change in share vs 2022 |
|---|---|---|---|
| Label-efficient and robust learning | 248 | 15.7% | +1.9 pts |
| Architectures and efficiency | 167 | 10.6% | -2.0 pts |
| Probabilistic and scientific modeling | 148 | 9.4% | -0.6 pts |
| Reinforcement learning | 148 | 9.4% | -2.0 pts |
| Learning theory and algorithms | 123 | 7.8% | -0.7 pts |
| Graph learning | 110 | 7.0% | -0.4 pts |
| Large language models | 82 | 5.2% | +2.9 pts |
| Optimization | 76 | 4.8% | -1.1 pts |
| Image generation | 75 | 4.8% | +0.6 pts |
| Adversarial robustness and security | 57 | 3.6% | -1.3 pts |
| Detection and segmentation | 53 | 3.4% | +0.2 pts |
| Vision-language and multimodal | 53 | 3.4% | +1.4 pts |
| Federated learning and privacy | 49 | 3.1% | +0.6 pts |
| 3D vision | 45 | 2.9% | +1.3 pts |
| Multi-agent systems and games | 37 | 2.3% | -0.7 pts |
| Continual learning | 27 | 1.7% | -0.5 pts |
| Time series | 23 | 1.5% | -0.8 pts |
| Video | 22 | 1.4% | +0.7 pts |
| Bandits and online learning | 21 | 1.3% | +0.4 pts |
| Point clouds | 11 | 0.7% | steady |
Fastest-growing topics at ICLR 2023
Change in share of ICLR papers from 2022 to 2023, topics with at least 10 papers.
- Language, safety and interpretability (Large language models): 17 → 46 papers
- Self-supervised and contrastive (Label-efficient and robust learning): 23 → 49 papers
- Agents, code and math (Large language models): 2 → 16 papers
- OOD and anomaly detection (Label-efficient and robust learning): 6 → 19 papers
- Cross-modal learning (Vision-language and multimodal): 2 → 11 papers
- Knowledge distillation (Label-efficient and robust learning): 4 → 13 papers
- NeRF and radiance fields (3D vision): 2 → 10 papers
- Personalized and heterogeneous FL (Federated learning and privacy): 9 → 20 papers
- Noisy labels (Label-efficient and robust learning): 5 → 14 papers
- CLIP and zero-shot (Vision-language and multimodal): 6 → 15 papers
More venues
Same year: ICML 2023 · NeurIPS 2023 · CVPR 2023 · AAAI 2023