Theoretical Understanding
Theoretical understanding in artificial intelligence currently focuses on rigorously analyzing the capabilities and limitations of various models, aiming to bridge the gap between empirical observations and formal guarantees. Research emphasizes developing theoretical frameworks for explaining model behavior, particularly in areas like large language models (LLMs), diffusion models, and graph neural networks, often employing techniques from information theory, optimization, and statistical learning theory to analyze model performance and generalization. These theoretical advancements are crucial for improving model design, enhancing reliability, and addressing concerns about robustness, fairness, and explainability, ultimately impacting the trustworthiness and responsible deployment of AI systems across diverse applications.
Papers
Mathematical Introduction to Deep Learning: Methods, Implementations, and Theory
Arnulf Jentzen, Benno Kuckuck, Philippe von Wurstemberger
Theory of Mind in Large Language Models: Examining Performance of 11 State-of-the-Art models vs. Children Aged 7-10 on Advanced Tests
Max J. van Duijn, Bram M. A. van Dijk, Tom Kouwenhoven, Werner de Valk, Marco R. Spruit, Peter van der Putten
Model-Based Reparameterization Policy Gradient Methods: Theory and Practical Algorithms
Shenao Zhang, Boyi Liu, Zhaoran Wang, Tuo Zhao
When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations
Aleksandar Petrov, Philip H. S. Torr, Adel Bibi
On the Theory of Risk-Aware Agents: Bridging Actor-Critic and Economics
Michal Nauman, Marek Cygan