This note outlines the primary options that can be adjusted to tune/improve results obtained with the Perple_X program MEEMUM. Additionally, it explains the auto-refine data prompt. The options initial_resolution, final_resolution, etc., are specified in a file named, by default, perplex_option.dat. These options are described in detail at www.perplex.ethz.ch/perplex_options.html. The problem definition file is created with the Perple_X program BUILD, it is named my_project.dat by default. Perple_X is a numerical method and should be tuned just as any other numerical method, i.e., the resolution parameters should be adjusted until there is no significant improvement in the results. The art of using Perple_X effectively is to not squander resources on insignificant improvement. 1) The prompt: Auto-refine data exists from a previous calculation with VERTEX. Do you want MEEMUM to use this data (y/n)? is issued if there is a file named my_project.arf in the local working directory. If you use this information MEEMUM runs more efficiently because it restricts the range of solution compositions considered during optimization. Additionally, the effective initial_resolution becomes initial_resolution/auto_refine_factor_I. This reduced (sic) initial_resolution increases quality and cost. Auto-refine data should not be used if the VERTEX calculation that generated the data did not span the entire range of physical conditions of interest for the calculations with MEEMUM. 2) initial_resolution: MEEMUM iteratively refines an initial optimization in which phase compositions are resolved at initial_resolution until a target resolution given by final_resolution is achieved. In principle, the iterative procedure provides high-quality results even if a large value of initial_resolution is specified. Phase separation is an exception to this principle, the resolution of solvi is tied to the initial_resolution value. To obtain high quality results, reduce initial_resolution until the results become insensitive to the specified value or MEEMUM runs out of memory (error ver041). It is relatively easy to find the smallest possible value of initial_resolution because the memory is allocated before any calculations, i.e., simply reduce initial_resolution until the program writes an error message before prompting for physical conditions. Using small values of initial resolution (~1/1000) is not costly for problems with only a few components, cost may be an issue for problems with a large number of components. 3) final_resolution: once the value of initial_resolution has been chosen, final_resolution can be adjusted to improve the accuracy of optimizations. final_resolution may be reduced until its value no longer affects the quality of the results or until it leads to numerical instability. final_resolution has no influence on memory usage and little cost in terms of computational time. 4) global_reach_increment/reach_increment: reach_increment is a means of increasing the numerical stability of the iterative stage of optimization. reach_increment expands the range of compositions considered during iteration. reach_increment is particularly useful for phases with flat Gibbs energy-composition surfaces (e.g., melts, alloys, and solvus phases), because the composition of such phases may change substantially during iteration. reach_increment may permit the use of smaller values of final_reolution, but this effect is generally not worth pursuing, i.e., my preferred strategy is to first optimize initial_resolution and final_resolution, and then apply reach_increment if necessary. reach_increment is computationally costly and may exhaust the allocated memory (error ver058). global_reach_increment applies reach_increment to all phases and is specified in the option file, reach_increment applies to a specific phase and is specified in the solution model file. 5) The speciation_factor and speciation_max_it option values may affect performance. My experience is that the default values are adequate for most purposes. 6) The default values of initial_resolution, final_resolution, and reach_increment are 1/16, 1e-3, and 0, respectively. I have used values of 1/1000, 1e-8, and 12 for calculations at high temperature. The improvement obtained by using such extreme values is usually only cosmetic, i.e., it is insignificant in view of the uncertainty in the thermodynamic models. 7) The influence of initial_resolution (controlled via its proxy auto_refine_factor_I), final_resolution, and reach_increment is illustrated more completely for calculations with VERTEX at www.perplex.ethz.ch/perplex_adaptive_min_keyword_group.html.