A new empirical sampling method termed "temperature intervals with global exchange of replicas and reduced radii" (TIGER3) is presented and demonstrated to efficiently equilibrate entangled long-chain molecular systems such as amorphous polymers. The TIGER3 algorithm is a replica exchange method in which simulations are run in parallel over a range of temperature levels at and above a designated baseline temperature. The replicas sampled at temperature levels above the baseline are run through a series of cycles with each cycle containing four stages: heating, sampling, quenching, and temperature level reassignment. The method allows chain segments to pass through one another at elevated temperature levels during the sampling stage by reducing the van der Waals radii of the atoms, thus eliminating chain entanglement problems. Atomic radii are then returned to their regular values and re-equilibrated at elevated temperature prior to quenching to the baseline temperature. Following quenching, replicas are compared using a Metropolis Monte Carlo exchange process for the construction of an approximate Boltzmann-weighted ensemble of states and then reassigned to the elevated temperature levels for additional sampling. Further system equilibration is performed by periodic implementation of the previously developed TIGER2 algorithm between cycles of TIGER3, which applies thermal cycling without radii reduction. When coupled with a coarse-grained modeling approach, the combined TIGER2/TIGER3 algorithm yields fast equilibration of bulk-phase models of amorphous polymer, even for polymers with complex, highly branched structures. The developed method was tested by modeling the polyethylene melt. The calculated properties of chain conformation and chain segment packing agreed well with published data. The method was also applied to generate equilibrated structural models of three increasingly complex amorphous polymer systems: poly(methyl methacrylate), poly(butyl methacrylate), and DTB succinate copolymer. Calculated glass transition temperature (;) and structural parameter profile (S(q)) for each resulting polymer model were found to be in close agreement with experimental T, values and structural measurements obtained by X-ray diffraction, thus validating that the developed methods provide realistic models of amorphous polymer structure.