Finetuning Foundation Models for Joint Analysis Optimization
This work addresses inefficiencies in HEP reconstruction and analysis for researchers in that field, representing an incremental improvement by applying existing ML techniques to a new domain.
The paper tackles the problem of sequential optimization in High Energy Physics (HEP) by proposing a joint approach using foundation models, resulting in significant gains in performance and data efficiency, as quantified in a search for heavy resonances decaying to four b-jets.
In this work we demonstrate that significant gains in performance and data efficiency can be achieved in High Energy Physics (HEP) by moving beyond the standard paradigm of sequential optimization or reconstruction and analysis components. We conceptually connect HEP reconstruction and analysis to modern machine learning workflows such as pretraining, finetuning, domain adaptation and high-dimensional embedding spaces and quantify the gains in the example usecase of searches of heavy resonances decaying via an intermediate di-Higgs system to four $b$-jets.