MergeVLA: Cross-Skill Model Merging Toward a Generalist Vision-Language-Action Agent Paper • 2511.18810 • Published Nov 24, 2025 • 1