The DNA of nuclear models: How AI predicts nuclear masses

Recently, many AI-based tools have shown promising results on this task, some achieving precision that surpasses the best physics models. However, the utility of these AI models remains in question given that predictions are only useful where measurements do not exist, which inherently requires extrapolation away from the training (and testing) samples. Since AI models are largely black boxes, the reliability of such an extrapolation is difficult to assess. For example, we find that (and explain why) the most important dimensions of its internal representation form a double helix, where the analog of the hydrogen bonds in DNA here link the number of protons and neutrons found in the most stable nucleus of each isotopic chain. Remarkably, the improvement of the AI model over symbolic ones can almost entirely be attributed to an observation made by Jaffe in 1969 based on the structure of most known nuclear ground states. The end result is a fully interpretable data-driven model of nuclear masses based on physics deduced by AI. Atomic nuclei consist of Z protons and N neutrons bound together by the strong nuclear force. Notably, many open problems in nuclear and (astro)particle physics are limited by a lack of precise knowledge of nuclear masses, either directly or indirectly via other quantities which require them as inputs. Experimentally, precise measurements have been made for the masses of (quasi)stable nuclei [9]; however, measurements of highly unstable nuclei are currently challenging, and thus, must be predicted using some combination of tractable theoretical calculations, e.g. using phenomeno-logical potentials, and empirical observations of other nuclei. Despite achieving an impressive level of precision, even the best such model is not sufficient to solve many open problems, e.g., r-process nucleosynthesis [10-12].