Symbolic regression for precision LHC physics

We explore how symbolic regression (SR) can be used to infer simple, accurate, and closed-form analytic expressions that can enhance the accuracy of phenomenological analyses at the Large Hadron Collider. We will conduct a thorough study by equation recovery in quantum electrodynamics processes, whose analytical descriptions are well-established from quantum field theory principles, to assess the method's validity. This will serve as a benchmark to evaluate the precision and robustness of SR before applying it to angular coefficients in $W$- and $Z$-boson production, a key process for Standard Model physics and new physics searches. In this way, SR combines the interpretability of traditional physics models with the power of machine learning techniques, offering the best of both worlds and facilitating scientific discovery.