ICML 2025 accepted papers by research area
3,342 papers were accepted at ICML 2025 (+732 vs 2024). Here is how they split across 20 research areas, and which topics gained the most ground.
See ICML 2025 in the interactive map
Research areas
| Area | Papers | Share | Change in share vs 2024 |
|---|---|---|---|
| Large language models | 538 | 16.1% | +4.2 pts |
| Probabilistic and scientific modeling | 341 | 10.2% | -2.1 pts |
| Learning theory and algorithms | 330 | 9.9% | -1.7 pts |
| Architectures and efficiency | 325 | 9.7% | +0.4 pts |
| Image generation | 237 | 7.1% | +1.9 pts |
| Label-efficient and robust learning | 226 | 6.8% | -1.6 pts |
| Reinforcement learning | 216 | 6.5% | -2.3 pts |
| Vision-language and multimodal | 178 | 5.3% | +1.3 pts |
| Graph learning | 175 | 5.2% | -0.2 pts |
| Optimization | 127 | 3.8% | -1.8 pts |
| Federated learning and privacy | 106 | 3.2% | -0.7 pts |
| Multi-agent systems and games | 97 | 2.9% | +0.9 pts |
| Bandits and online learning | 96 | 2.9% | -0.2 pts |
| Adversarial robustness and security | 89 | 2.7% | +0.2 pts |
| Time series | 69 | 2.1% | steady |
| Continual learning | 51 | 1.5% | +0.6 pts |
| Video | 48 | 1.4% | +0.6 pts |
| 3D vision | 40 | 1.2% | +0.1 pts |
| Detection and segmentation | 34 | 1.0% | +0.1 pts |
| Point clouds | 19 | 0.6% | +0.1 pts |
Fastest-growing topics at ICML 2025
Change in share of ICML papers from 2024 to 2025, topics with at least 10 papers.
- Fast diffusion sampling (Image generation): 48 → 103 papers
- Language, safety and interpretability (Large language models): 100 → 169 papers
- Reasoning and chain of thought (Large language models): 33 → 77 papers
- Multimodal LLMs (Vision-language and multimodal): 20 → 57 papers
- Agents, code and math (Large language models): 46 → 89 papers
- Autoregressive and latent generation (Image generation): 39 → 72 papers
- LLM agents (Multi-agent systems and games): 15 → 40 papers
- Alignment and preferences (Large language models): 32 → 58 papers
- Molecules and proteins (Probabilistic and scientific modeling): 69 → 104 papers
- Graph self-supervised learning (Graph learning): 38 → 63 papers
More venues
Same year: NeurIPS 2025 · ICLR 2025 · CVPR 2025 · AAAI 2025