T. R. Harvey, A. Lukas

In this paper, we apply reinforcement learning to particle physics model building. As an example environment, we use the space of Froggatt-Nielsen type models for quark masses. Using a basic policy-based algorithm we show that neural networks can be successfully trained to construct Froggatt-Nielsen models which are consistent with the observed quark masses and mixing. The trained policy networks lead from random to phenomenologically acceptable models for over 90% of episodes and after an average episode length of about 20 steps. We also show that the networks are capable of finding models proposed in the literature when starting at nearby configurations.

Alfredo Guevara, Elizabeth Himwich, Monica Pate, Andrew Strominger

All 4D gauge and gravitational theories in asymptotically flat spacetimes contain an infinite number of non-trivial symmetries. They can be succinctly characterized by generalized 2D currents acting on the celestial sphere. A complete classification of these symmetries and their algebras is an open problem. Here we construct two towers of such 2D currents from positive-helicity photons, gluons, or gravitons with integer conformal weights $k=1,0,-1,...$. These generate the symmetries associated to an infinite tower of conformally soft theorems. The current algebra commutators are explicitly derived from the poles in the OPE coefficients, and found to comprise a rich closed subalgebra of the complete symmetry algebra.

Gerard t Hooft

The machinery of quantum mechanics is fully capable of describing a single ontological world. Here we discuss the converse: in spite of appearances, and indeed numerous claims to the contrary, any quantum mechanical model can be mimicked, up to any required accuracy, by a completely classical system of equations. An implication of this observation is that Bell's theorem cannot hold in many cases. This is explained by scrutinising Bell's assumptions concerning causality, retrocausality, statistical (in-)dependence, and his fear of `conspiracy' (there is no conspiracy in our constructions). The potential importance of our construction in model building is discussed.

Andrew G. Cohen, David B. Kaplan

In a previous paper, the authors with Ann Nelson proposed that the UV and IR applicability of effective quantum field theories should be constrained by requiring that strong gravitational effects are nowhere encountered in a theory's domain of validity [Phys. Rev. Lett. 82, 4971 (1999)]. The constraint was proposed to delineate the boundary beyond which conventional quantum field theory, viewed as an effective theory excluding quantum gravitational effects, might be expected to break down. In this Letter we revisit this idea and show that quantum gravitational effects could lead to a deviation of size $(\alpha/2\pi)\sqrt{m_e/M_p}$ from the Standard Model calculation for the electron magnetic moment. This is the same size as QED and hadronic uncertainties in the theory of $a_e$, and a little more than one order of magnitude smaller than both the dominant uncertainty in its Standard Model value arising from the accuracy with which $\alpha$ is measured, as well as the experimental uncertainty in measurement of $a_e$.