NeurIPS 2024 accepted papers by research area
4,560 papers were accepted at NeurIPS 2024 (+976 vs 2023). Here is how they split across 20 research areas, and which topics gained the most ground.
See NeurIPS 2024 in the interactive map
Research areas
| Area | Papers | Share | Change in share vs 2023 |
|---|---|---|---|
| Large language models | 612 | 13.4% | +7.4 pts |
| Learning theory and algorithms | 402 | 8.8% | -3.1 pts |
| Probabilistic and scientific modeling | 390 | 8.6% | -2.1 pts |
| Architectures and efficiency | 388 | 8.5% | +0.8 pts |
| Label-efficient and robust learning | 338 | 7.4% | -2.1 pts |
| Image generation | 333 | 7.3% | +0.2 pts |
| Vision-language and multimodal | 318 | 7.0% | +2.5 pts |
| Reinforcement learning | 271 | 5.9% | -1.7 pts |
| 3D vision | 217 | 4.8% | +1.4 pts |
| Optimization | 191 | 4.2% | -0.7 pts |
| Graph learning | 184 | 4.0% | -0.9 pts |
| Video | 143 | 3.1% | +0.8 pts |
| Bandits and online learning | 128 | 2.8% | -0.8 pts |
| Detection and segmentation | 126 | 2.8% | -0.2 pts |
| Multi-agent systems and games | 119 | 2.6% | +0.2 pts |
| Federated learning and privacy | 114 | 2.5% | -1.1 pts |
| Adversarial robustness and security | 101 | 2.2% | -1.0 pts |
| Time series | 85 | 1.9% | +0.4 pts |
| Point clouds | 52 | 1.1% | -0.1 pts |
| Continual learning | 48 | 1.1% | +0.2 pts |
Fastest-growing topics at NeurIPS 2024
Change in share of NeurIPS papers from 2023 to 2024, topics with at least 10 papers.
- Language, safety and interpretability (Large language models): 82 → 202 papers
- Multimodal LLMs (Vision-language and multimodal): 7 → 97 papers
- Agents, code and math (Large language models): 39 → 131 papers
- Gaussian splatting (3D vision): 0 → 65 papers
- Instruction and fine-tuning (Large language models): 17 → 77 papers
- Attention and transformers (Architectures and efficiency): 65 → 130 papers
- Alignment and preferences (Large language models): 11 → 56 papers
- Video generation (Video): 11 → 55 papers
- Retrieval and knowledge (Large language models): 24 → 71 papers
- 3D generation (3D vision): 34 → 68 papers