Improving the matrix multiplication exponent with modern optimization and AlphaEvolve
The paper lowers the best known upper bound for the matrix multiplication exponent to $\omega < 2.371177$.
The authors target the optimization problem inside combination loss analysis, the refinement behind the current best bounds. They say they reformulate that problem so it can be solved in a larger setting than before. They then build a new optimization algorithm using recent machine-learning advances and refine it with AlphaEvolve. The previous best bound cited was 2.371339. HF Daily Papers' note
The authors target the optimization problem inside combination loss analysis, the refinement behind the current best bounds. They say they reformulate that problem so it can be solved in a larger setting than before. They then build a new optimization algorithm using recent machine-learning advances and refine it with AlphaEvolve. The previous best bound cited was 2.371339. HF Daily Papers' note
score 7