This paper introduces a generative diffusion model to discover new periodic orbits in 2D Navier-Stokes equations, even when trained on non-periodic turbulent data. By modifying the model and enforcing symmetries, it generated plausible candidates which were then refined into 111 previously unknown exact solutions, highlighting generative AI's role as a complementary tool for exploring complex solution spaces.

This paper introduces a resampling technique for trained AI models that leverages invariant transformations of input data to reduce epistemic uncertainty and improve inference accuracy. By aggregating inferences from multiple transformed samples, the method offers a way to enhance performance without re-training, potentially balancing model size and effectiveness.

This paper proposes an agentic AI framework for intent-driven optimization in cell-free Open Radio Access Networks (O-RAN), where LLM-based agents collaborate to translate operator intents into network optimizations. The framework demonstrates significant reductions in active O-RUs for energy saving and memory usage through parameter-efficient fine-tuning.

The paper explores human-AI collaboration in mathematical research through a case study on Hermite quadrature error estimation, demonstrating AI's capabilities in algebraic manipulation and proof exploration while emphasizing the critical need for human verification and strategic direction to mitigate errors and ensure novel discovery.

This paper introduces a theoretical framework for accelerating the evaluation of Conditional Value-at-Risk (CVaR) value functions in Partially Observable Markov Decision Processes (POMDPs) with formal performance guarantees. It derives novel CVaR bounds for random variables, enabling faster policy evaluation through action elimination using simplified models.