Mass Agnostic Jet Assignment using Symmetry Preserving Attention Networks.

Discovering new resonant particles at the CERN LHC often requires analyzing multijet final states where the mass and decay topology of the unknown resonance are a priori undetermined. Traditional searches rely on fixed reconstruction assumptions, limiting their sensitivity across a broad invariant mass range. Building on our previous development of symmetry preserving attention network (SPA-Net) for multi-Higgs reconstruction, we extend the approach to perform mass agnostic searches for new resonances that may appear in any combination of boosted and resolved topologies. The model simultaneously reconstructs candidate particles from small and large radius jets, automatically adapting to the kinematics of each event. This unified reconstruction framework enables efficient and model independent searches for resonant structures without assuming a specific mass hypothesis, providing a new avenue for discovering physics beyond the standard model.