Search papers, labs, and topics across Lattice.
This paper refines the optimization problem central to the combination loss analysis method for improving the matrix multiplication exponent $\omega$. By reformulating the optimization landscape and integrating a novel machine learning-based algorithm enhanced with AlphaEvolve, the authors achieve a new upper bound of $\omega < 2.371177$. This result surpasses the previous best bound of 2.371339, marking a significant advancement in the field of matrix multiplication algorithms.
Achieving a new upper bound of $\omega < 2.371177 could redefine our understanding of matrix multiplication efficiency.
The current best bounds on the matrix multiplication exponent $\omega$ are obtained through a refinement of the laser method called combination loss analysis (Duan et al., 2022; Williams et al., 2024; Alman et al., 2025). In this note, we address the optimization problem at the core of this approach and propose several improvements. First, we reformulate the optimization problem allowing us to solve it in a larger setting than was previously possible. Second, we leverage recent advances in machine learning to design a new optimization algorithm for this problem. Finally, we refine the resulting optimization algorithm with AlphaEvolve. Our combined approach yields an upper bound of $\omega$<2.371177, improving the previous best bound of 2.371339.
CMU ML