Demystifying Mergeability: Interpretable Properties to Predict Model Merging Success

Model merging combines knowledge from separately fine-tuned models, yet the factors driving its success remain poorly understood. While recent work treats mergeability as an intrinsic property of the models, we show with an architectureagnostic framework that it fundamentally depends on both the merging method and the partner tasks. Using L1-regularized linear optimization over a set of interpretable pairwise metrics (e.g., gradient L 2 distance), we uncover properties correlating with post-merge normalized accuracy across five merging methods. We find that the drivers of merge success vary across architectures and merging methods, overall with only a moderate agreement (64.0% average top-5 metric overlap; 79.3% sign agreement). Crucially, however, gradient alignment metrics consistently emerge as the most fundamental signals of mergeability. These findings provide a diagnostic foundation for understanding mergeability and motivate future mergeaware fine-tuning strategies. Despite its practical appeal, model merging remains highly unpredictable. Some model pairs merge seamlessly, yielding performance close to the average of their individual task