Recent research in AI theory is increasingly focused on understanding the fundamental principles that govern intelligence and agency, particularly in complex systems. A notable trend is the exploration of how AI can effectively manage uncertainty, with distinctions made between epistemic and subjective uncertainty, which could enhance decision-making processes in uncertain environments. Additionally, investigations into the expressive power of transformer architectures are revealing their capabilities in approximating complex functions, which has implications for optimizing model performance across various applications. Theoretical advancements are also being made in Bayesian network learning, with new parameterization strategies that could simplify the learning process. Furthermore, the analysis of reasoning mechanisms in large language models is shedding light on the computational costs associated with chain-of-thought reasoning, providing insights into optimizing inference time. Collectively, these developments are paving the way for more resilient, adaptive AI systems that can better navigate real-world complexities and uncertainties.
To operate reliably under changing conditions, complex systems require feedback on how effectively they use resources, not just whether objectives are met. Current AI systems process vast information ...
Transformer networks have achieved remarkable empirical success across a wide range of applications, yet their theoretical expressive power remains insufficiently understood. In this paper, we study t...
The paper investigates whether and how AI systems can realize states of uncertainty. By adopting a functionalist and behavioral perspective, it examines how symbolic, connectionist and hybrid architec...
Understanding why gradient-based training in deep networks exhibits strong implicit bias remains challenging, in part because tractable singular-value dynamics are typically available only for balance...
Deep neural networks (DNNs) have achieved remarkable empirical success, yet the absence of a principled theoretical foundation continues to hinder their systematic development. In this survey, we pres...
Foundation models excel in stable environments, yet often fail where reliability matters most: medicine, finance, and policy. This Fidelity Paradox is not just a data problem; it is structural. In dom...
We investigate the parameterized complexity of Bayesian Network Structure Learning (BNSL), a classical problem that has received significant attention in empirical but also purely theoretical studies....
Different types of reasoning impose different structural demands on representational systems, yet no systematic account of these demands exists across psychology, AI, and philosophy of mind. I propose...
Analogy is a central faculty of human intelligence, enabling abstract patterns discovered in one domain to be applied to another. Despite its central role in cognition, the mechanisms by which Transfo...
When designing compound AI systems, a common approach is to query multiple copies of the same model and aggregate the responses to produce a synthesized output. Given the homogeneity of these models, ...