AAAI 2023 accepted papers by research area

2,023 papers were accepted at AAAI 2023 (+704 vs 2022). 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 2022
Label-efficient and robust learning	345	17.1%	+1.9 pts
Learning theory and algorithms	229	11.3%	-3.6 pts
Graph learning	173	8.6%	+0.5 pts
Probabilistic and scientific modeling	129	6.4%	+0.7 pts
Architectures and efficiency	128	6.3%	-0.3 pts
Large language models	125	6.2%	+1.6 pts
Reinforcement learning	114	5.6%	-0.3 pts
Multi-agent systems and games	102	5.0%	+0.3 pts
Detection and segmentation	100	4.9%	-1.3 pts
Vision-language and multimodal	88	4.3%	-0.2 pts
Image generation	85	4.2%	+0.5 pts
Video	81	4.0%	steady
Adversarial robustness and security	71	3.5%	steady
3D vision	49	2.4%	-0.2 pts
Federated learning and privacy	43	2.1%	steady
Time series	37	1.8%	+0.8 pts
Bandits and online learning	35	1.7%	-0.4 pts
Point clouds	31	1.5%	-0.3 pts
Continual learning	30	1.5%	+1.0 pts
Optimization	28	1.4%	-0.8 pts

Change in share of AAAI papers from 2022 to 2023, topics with at least 10 papers.

AI and society (Large language models): 2 → 52 papers
Recognition and long-tail (Label-efficient and robust learning): 98 → 186 papers
Graph self-supervised learning (Graph learning): 23 → 52 papers
Online and replay-based (Continual learning): 4 → 20 papers
Human-AI and goal-conditioned (Multi-agent systems and games): 27 → 55 papers
Privacy attacks and secure learning (Federated learning and privacy): 2 → 15 papers
Uncertainty and Gaussian processes (Probabilistic and scientific modeling): 6 → 21 papers
Knowledge and temporal graphs (Graph learning): 34 → 63 papers
RNNs and normalization (Architectures and efficiency): 24 → 46 papers
Causal and generative time series (Time series): 6 → 17 papers

AAAI in other years: 2024 · 2025 · 2026