Mapping phase diagrams has recently been improved through active learning frameworks that prioritize sampling informative points over simple grid searches. The common Bayesian optimization framework with a Gaussian process classifier initially samples a small set of points and then uses an acquisition function to select the next point to sample. Although this approach improves efficiency by making informed decisions at every iteration, sampling only one point at a time remains experimentally tedious. The acquisition function determines the informativeness of every point in the sampling space. Individual sampling selects the point with the highest acquisition value to be sampled next. Batch sampling methods optimize this process by selecting various points that satisfy distance or acquisition value thresholds at each iteration. Three batch sampling strategies were tested: fixed batch size, ranged-fixed batch with a distance constraint around each sampled point, and an adaptable batch size with a minimum acquisition value constraint. These methods were evaluated on synthetic phase diagrams of varying complexity and compared against individual sampling. The ranged-fixed batch method, with a batch size of 10 and radius of 0.2 (2/N), consistently produced phase diagrams with higher accuracy in fewer iterations and comparable total sample sizes relative to both individual sampling and the other batch methods. These results highlight batch sampling as a practical improvement for phase diagram mapping and a step toward streamlining experimental workflows.