CVPR 2025 accepted papers by research area

2,871 papers were accepted at CVPR 2025 (+155 vs 2024). Here is how they split across 20 research areas, and which topics gained the most ground.

Research areas

Area	Papers	Share	Change in share vs 2024
3D vision	602	21.0%	-2.3 pts
Image generation	481	16.8%	+0.2 pts
Vision-language and multimodal	448	15.6%	+3.5 pts
Video	391	13.6%	+3.6 pts
Label-efficient and robust learning	222	7.7%	-1.1 pts
Detection and segmentation	198	6.9%	-4.0 pts
Architectures and efficiency	144	5.0%	+0.6 pts
Point clouds	103	3.6%	-0.2 pts
Adversarial robustness and security	63	2.2%	-0.1 pts
Continual learning	50	1.7%	steady
Large language models	37	1.3%	+0.4 pts
Multi-agent systems and games	27	0.9%	steady
Federated learning and privacy	26	0.9%	-0.5 pts
Graph learning	25	0.9%	+0.2 pts
Probabilistic and scientific modeling	20	0.7%	-0.3 pts
Learning theory and algorithms	15	0.5%	-0.1 pts
Reinforcement learning	10	0.3%	steady
Optimization	5	0.2%	steady
Time series	3	0.1%	steady
Bandits and online learning	1	0.0%	steady

Change in share of CVPR papers from 2024 to 2025, topics with at least 10 papers.

Multimodal LLMs (Vision-language and multimodal): 56 → 154 papers
Gaussian splatting (3D vision): 52 → 147 papers
Video generation (Video): 66 → 157 papers
Video-language understanding (Video): 62 → 111 papers
Autoregressive and latent generation (Image generation): 74 → 103 papers
Fast diffusion sampling (Image generation): 57 → 77 papers
Semi-supervised learning (Label-efficient and robust learning): 13 → 26 papers
3D generation (3D vision): 169 → 188 papers
Agents, code and math (Large language models): 7 → 16 papers
Attacks on text and graphs (Adversarial robustness and security): 25 → 34 papers

CVPR in other years: 2023 · 2024 · 2026