A set of Monte Carlo simulated events, for the evaluation of top quarks' (and their child particles') momentum reconstruction. Produced using the HEPData4ML package [1]. The data is saved in HDF5 format, as sets of arrays with keys (as detailed below). There are ~1.15M events, with approximately 700k in "train.h5", 200k in "valid.h5", and 250k in "test.h5". There are two versions of the data, the difference between them being whether or not (fast) detector simulation was performed. Those with the detector simulation have the "_delphes" suffix in their filenames. Both versions are produced from the same set of generator-level events. 13 TeV center-of-mass energy, fully hadronic top quark decays, simulated with Pythia8. Events are generated with leading top quark pT in [550,650] GeV. Where applicable, detector simulation is done using Delphes, with the ATLAS detector card. Clustering of particles/objects is done using the anti-kT algorithm, with \(R=0.8\). For the data without detector simulation, the inputs to clustering are the stable, visible final-state particles from Pythia8. For the data with detector simulation, the inputs are calorimeter towers (`Towers`) from Delphes. Each entry corresponds with a single jet. All jets are matched to a parton-level top quark within \(\Delta R =0.8\) Jets are required to have \(|\eta| < 2, \; p_T > 15 \text{ GeV}\) The 200 leading (highest \(p_T\)) jet constituent four-momenta are stored in Cartesian coordinates \((E,p_x,p_y,p_z)\), sorted by decreasing \(p_T\) and with with zero-padding for jets with fewer than 200 constituents. These are stored under the key `Pmu`. The number of non-zero jet constituents is stored under the key `Nobj`. The jet four-momentum is stored in Cartesian coordinates and in cylindrical coordinates \((p_T,\eta,\phi,m)\) under keys `jet_Pmu` and `jet_Pmu_cyl`, respectively. The truth (parton-level) four-momenta of the top quark, and the bottom quark and W-boson to which it decays, are stored in Cartesian coordinates in keys `truth_Pmu_0`, `truth_Pmu_1` and `truth_Pmu_2` respectively. In addition, these are stored together under the key `truth_Pmu`, with the corresponding PDG codes stored under the key `truth_Pdg`. The testing files `test.h5` and `test_delphes.h5` contain additional information for gauging performance, primarily on measurements of the W boson: In addition to the parton-level top, bottom and W-boson, the truth-level particles in `truth_Pmu` also include the two quarks from W decay, followed by up to 200 stable daughter particles from the W boson decay. `is_W_daughter`: An array of {0,1} indicating whether or not a particular jet constituent in `Pmu` is matched to a truth-level W daughter. This particular array is only present in the testing file without detector simulation. `jet_is_contained`: Whether or not one of the quarks from W decay has a distance from the jet center of \(\Delta R > 0.8\) . `jet_q_dr_max`: The maximum distance between the jet center and one of the quarks from W decay in \((\eta,\phi)\). `jh_tag`: Whether or not this jet was tagged by the Johns Hopkins top tagger [2], as implemented in Fastjet. `jh_W_pred`: The four-momentum of the W boson candidate identified by the JH tagger (only present for tagged jets), in Cartesian coordinates. `jh_W_pred_constituents`: Up to 200 constituent four-momenta of the JH W boson candidate, in Cartesian coordinates. `jh_W_nobj`: The number of constituent four-momenta of the JH W boson candidate. `is_jh_constituent`: An array of {0,1} indicating whether or not a particular jet constituent in `Pmu` is matched to a JH W boson candidate constituent. `jh_m_pred`: The mass of the JH W boson candidate. `jh_pt_pred`: The \(p_T\) of the JH W boson candidate. `jh_m_res`: The ratio of the JH W boson candidate mass to the true W boson mass. `jh_pt_res`: The ratio of the JH W boson candidate \(p_T\) to the true W boson \(p_T\). `jh_psi`: The lab-frame angle between the JH W boson candidate 3-momentum and the true W boson 3-momentum. `event_idx`: An integer indexing the event number, which may be useful for bookkeeping if splitting the testing file. `process_code`: The process code as given by Pythia. `cross_section`: The cross section estimate for the process that produced this jet, as given by Pythia. Potentially useful if combining this dataset with other datasets that involve different processes. `cross_section_uncertainty`: The cross section estimate uncertainty as given by Pythia. `mc_weight`: The generator weight for each event as given by Pythia. [1]: J. T. Offermann and X. Liu, HEPData4ML, (2022). [2]: D. E. Kaplan, K. Rehermann, M. D. Schwartz, and B. Tweedie, Top Tagging: A Method for Identifying Boosted Hadronically Decaying Top Quarks, Phys. Rev. Lett. 101, 142001 (2008).