2026 edition
The best papers in each domain
12 pages, each with sub-sections. A paper is ranked on its main domain only and shown as “also relevant” on the others. The length of each list scales with the size of its field.
Top 80
Machine learning foundations
Learning theory, statistics, optimisation and probabilistic modelling: the ideas the rest of the field stands on.
Sections: Learning theory & classic ML, Probabilistic ML · #1: Maximum Likelihood from Incomplete Data Via the EM Algorithm
Top 65
Deep learning
Neural network architectures, training methods, and the systems that make them run efficiently.
Sections: Architectures & training, Efficient ML & systems · #1: Adam: A Method for Stochastic Optimization
Top 85
Computer vision & multimodal
Seeing machines: recognition, detection, segmentation, self-supervised visual features and vision-language models.
Sections: Computer vision, Multimodal & vision-language, Self-supervised learning · #1: ImageNet Classification with Deep Convolutional Neural Networks
Top 100
Language, speech & LLMs
Natural language processing, large language models, and speech and audio.
Sections: Large language models, Natural language processing, Speech & audio · #1: Neural Machine Translation by Jointly Learning to Align and Translate
Top 35
Reinforcement learning
Learning to act from reward: dynamic programming, value and policy methods, and deep RL.
Sections: Reinforcement learning · #1: Continuous control with deep reinforcement learning
Top 30
Generative models
Models that learn to produce data: VAEs, GANs, flows, diffusion and autoregressive models.
Sections: Generative models · #1: Auto-Encoding Variational Bayes
Top 30
Graph learning
Learning on graphs and relational data: embeddings, graph neural networks, geometric deep learning.
Sections: Graph learning · #1: Semi-Supervised Classification with Graph Convolutional Networks
Top 35
Robotics
Perception, planning and control for machines in the physical world.
Sections: Robotics · #1: Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
Top 40
Classic AI: search, logic & planning
Symbolic AI from 1943 on: search, logic, knowledge representation, planning and reasoning under uncertainty.
Sections: Classic AI · #1: A Formal Basis for the Heuristic Determination of Minimum Cost Paths
Top 35
Search, retrieval & recommendation
Information retrieval, web search, ranking and recommender systems.
Sections: Retrieval & recommendation · #1: The Anatomy of a Large-Scale Hypertextual Web Search Engine
Top 70
Safety, alignment & interpretability
Making AI systems understandable, robust and aligned with what people want.
Sections: Interpretability, Safety & alignment · #1: Intriguing properties of neural networks
Top 35
AI for science
AI applied to the sciences: proteins, materials, weather, mathematics and medicine.
Sections: AI for science · #1: Highly accurate protein structure prediction with AlphaFold
Data sources
Catalogue data from OpenAlex (CC0), Crossref (CC0 metadata; no abstracts are shown), arXiv, and Papers With Code (CC BY-SA 4.0). Each count names its source. See the method.