1: \begin{abstract}

2:     We develop a framework for proving rapid convergence of shifted QR algorithms which use  Ritz values as shifts, in finite arithmetic. Our key contribution is a dichotomy result which addresses the known forward-instability issues surrounding the shifted QR iteration \cite{parlett1993forward}: we give a procedure which provably {\em either} computes a set of approximate Ritz values of a Hessenberg matrix with good forward stability properties, {\em or} leads to early decoupling of the matrix via a small number of QR steps.

3:     

4:     Using this framework, we show that the shifting strategy of \cite{banks2021global} converges rapidly in finite arithmetic with a polylogarithmic bound on the number of bits of precision required, when invoked on matrices of controlled eigenvector condition number and minimum eigenvalue gap. 

5:     

6: \end{abstract}

7: