The Large Hadron Collider provides proton-proton collisions at the highest energies ever achieved in a laboratory setting, enabling precision tests of the Standard Model and searches for new physics. Many of the resulting final states are hadronic and are reconstructed as jets, whose measurement and interpretation are central to a broad range of analyses. This thesis presents a novel methodology to access an unprecedented dataset of low-energy hadronic interactions in ATLAS by exploiting pile-up collisions in Run 2. Instead of analyzing only the highest-energy interaction in each bunch crossing, the triggering interaction is vetoed and the remaining pile-up interactions are used, yielding a large, trigger-unbiased sample dominated by low-energy jet production.

The utility of the pile-up dataset is demonstrated through a measurement of the jet energy resolution using the dijet balance technique, extending sensitivity well below the energies accessible with standard jet-trigger data and providing an independent validation of current low-energy calibration strategies. After accounting for analysis selections, the pile-up dataset achieves superior statistical precision for jet transverse momenta below 65 GeV, providing up to 50 times more events than the single-jet triggers for this measurement. The pile-up dataset is further utilized in a search for low-mass dijet resonances, constituting the first-ever search for new physics in pile-up collisions. The search is performed in the mass range 100-250 GeV for both model-agnostic Gaussian resonances and a benchmark model of an axial-vector dark matter mediator coupling to quarks. No significant excess is found from the background expectation, and exclusion limits on dijet resonance cross-sections are derived for both models. These limits represent the lowest-energy resolved dijet mass exclusion limits set at the LHC, demonstrating the unique sensitivity of this dataset in searches for physics beyond the Standard Model.