NeurIPS 2025 accepted papers by research area
5,812 papers were accepted at NeurIPS 2025 (+1,252 vs 2024). Here is how they split across 20 research areas, and which topics gained the most ground.
See NeurIPS 2025 in the interactive map
Research areas
| Area | Papers | Share | Change in share vs 2024 |
|---|---|---|---|
| Large language models | 962 | 16.6% | +3.1 pts |
| Vision-language and multimodal | 539 | 9.3% | +2.3 pts |
| Probabilistic and scientific modeling | 521 | 9.0% | +0.4 pts |
| Architectures and efficiency | 512 | 8.8% | +0.3 pts |
| Image generation | 440 | 7.6% | +0.3 pts |
| Learning theory and algorithms | 405 | 7.0% | -1.8 pts |
| Label-efficient and robust learning | 345 | 5.9% | -1.5 pts |
| Reinforcement learning | 323 | 5.6% | -0.4 pts |
| 3D vision | 265 | 4.6% | -0.2 pts |
| Video | 235 | 4.0% | +0.9 pts |
| Multi-agent systems and games | 213 | 3.7% | +1.1 pts |
| Graph learning | 202 | 3.5% | -0.6 pts |
| Optimization | 186 | 3.2% | -1.0 pts |
| Bandits and online learning | 130 | 2.2% | -0.6 pts |
| Federated learning and privacy | 120 | 2.1% | -0.4 pts |
| Detection and segmentation | 113 | 1.9% | -0.8 pts |
| Adversarial robustness and security | 99 | 1.7% | -0.5 pts |
| Time series | 99 | 1.7% | -0.2 pts |
| Continual learning | 55 | 0.9% | -0.1 pts |
| Point clouds | 48 | 0.8% | -0.3 pts |
Fastest-growing topics at NeurIPS 2025
Change in share of NeurIPS papers from 2024 to 2025, topics with at least 10 papers.
- Reasoning and chain of thought (Large language models): 41 → 225 papers
- Multimodal LLMs (Vision-language and multimodal): 97 → 260 papers
- LLM agents (Multi-agent systems and games): 30 → 105 papers
- RL from feedback (Reinforcement learning): 21 → 80 papers
- Video-language understanding (Video): 33 → 83 papers
- Molecules and proteins (Probabilistic and scientific modeling): 99 → 163 papers
- Fast diffusion sampling (Image generation): 107 → 168 papers
- Robot action and navigation (Vision-language and multimodal): 14 → 49 papers
- Agents, code and math (Large language models): 131 → 198 papers
- Video generation (Video): 55 → 92 papers