Scattering amplitudes at high loop orders are remarkably difficult for humans to compute. Can machines do any better? As a first exercise, we map a set of scattering amplitudes into a “language-like” representation using the symbol associated with multiple polylogarithms. Then we train a transformer-based model (think ChatGPT) to predict (integer) coefficients of “words” in the symbol. Such models can also learn correlations between coefficients at different loop orders. I also discuss the next phase(s) of this work, i.e. whether one can predict the next loop order from information gleaned at lower orders.

SLAC