ICLR 2026 accepted papers by research area
5,358 papers were accepted at ICLR 2026 (+1,655 vs 2025). Here is how they split across 20 research areas, and which topics gained the most ground.
See ICLR 2026 in the interactive map
Research areas
| Area | Papers | Share | Change in share vs 2025 |
|---|---|---|---|
| Large language models | 1,219 | 22.8% | +2.7 pts |
| Vision-language and multimodal | 555 | 10.4% | +3.1 pts |
| Image generation | 476 | 8.9% | -0.4 pts |
| Architectures and efficiency | 464 | 8.7% | -1.6 pts |
| Reinforcement learning | 390 | 7.3% | +1.5 pts |
| Probabilistic and scientific modeling | 315 | 5.9% | -1.9 pts |
| Label-efficient and robust learning | 259 | 4.8% | -1.4 pts |
| Multi-agent systems and games | 256 | 4.8% | +1.6 pts |
| Video | 247 | 4.6% | +0.6 pts |
| Learning theory and algorithms | 218 | 4.1% | -0.7 pts |
| 3D vision | 208 | 3.9% | +0.3 pts |
| Graph learning | 144 | 2.7% | -1.4 pts |
| Optimization | 116 | 2.2% | -1.2 pts |
| Time series | 110 | 2.1% | +0.3 pts |
| Adversarial robustness and security | 104 | 1.9% | -0.7 pts |
| Federated learning and privacy | 81 | 1.5% | -0.2 pts |
| Detection and segmentation | 63 | 1.2% | -0.4 pts |
| Continual learning | 58 | 1.1% | steady |
| Bandits and online learning | 52 | 1.0% | steady |
| Point clouds | 23 | 0.4% | -0.2 pts |
Fastest-growing topics at ICLR 2026
Change in share of ICLR papers from 2025 to 2026, topics with at least 10 papers.
- Reasoning and chain of thought (Large language models): 67 → 347 papers
- Multimodal LLMs (Vision-language and multimodal): 132 → 323 papers
- LLM agents (Multi-agent systems and games): 59 → 197 papers
- RL from feedback (Reinforcement learning): 28 → 134 papers
- Robot action and navigation (Vision-language and multimodal): 25 → 76 papers
- Autoregressive and latent generation (Image generation): 57 → 107 papers
- Cross-modal learning (Vision-language and multimodal): 24 → 59 papers
- Agents, code and math (Large language models): 134 → 213 papers
- Super-resolution (Image generation): 2 → 21 papers
- Video generation (Video): 82 → 135 papers
More venues
Same year: ICML 2026 · NeurIPS 2026 · CVPR 2026 · AAAI 2026