Global Bayesian Synthesis of Conflicting BSM Anomalies with Dynamic Model Updating

Papers like Barducci et al. (2025) and Arnan et al. (2017) re-examine anomalies as new experimental results emerge, but typically do so in a piecemeal, model-focused way. My idea is to construct a global Bayesian inference framework that can ingest new experimental data (including null results or conflicting findings) and dynamically update the posterior probabilities of entire classes of BSM models—reweighting, combining, or even discarding models as the landscape shifts. This approach would use multi-objective ML parameter scans (as in Diaz et al. 2024) and explicit model selection criteria, making the search for new physics more robust to conflicting or evolving experimental evidence. The innovation lies in treating the anomalous landscape as a living, probabilistic ensemble, rather than a static set of puzzles, enabling real-time theory-experiment interplay.

References:

1. On the Atomki nuclear anomaly after the MEG-II result. D. Barducci, Davide Germani, M. Nardecchia, Stefano Scacco, C. Toni (2025). Journal of High Energy Physics.
2. Loop effects of heavy new scalars and fermions in b → sμ+μ−. P. Arnan, A. Crivellin, L. Hofer, F. Mescia (2017). Journal of High Energy Physics.
3. hep-aid: A Python Library for Sample Efficient Parameter Scans in Beyond the Standard Model Phenomenology. Mauricio A. Diaz, S. Dasmahapatra, Stefano Moretti (2024).