AI RESEARCH
Lance: Unified Multimodal Modeling by Multi-Task Synergy
arXiv CS.CV
•
ArXi:2605.18678v1 Announce Type: new We present Lance, a lightweight native unified model ing multimodal understanding, generation, and editing for both images and videos. Rather than relying on model capacity scaling or text-image-dominant designs, Lance explores a practical paradigm for unified multimodal modeling via collaborative multi-task