Lance: Unified Multimodal Modeling by Multi-Task Synergy

TL;DR

Lance is a lightweight, unified multimodal model that excels in understanding, generating, and editing images and videos through collaborative multi-task training and a dual-stream architecture.

Contribution

It introduces a practical paradigm for unified multimodal modeling with a dual-stream mixture-of-experts architecture and modality-aware positional encoding, improving performance across tasks.

Findings

Lance outperforms existing open-source models in image and video generation.

It maintains strong multimodal understanding capabilities.

Lance's staged multi-task training enhances semantic and visual performance.

Abstract

We present Lance, a lightweight native unified model supporting 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 training. It is grounded in two core principles: unified context modeling and decoupled capability pathways. Specifically, Lance is trained from scratch and employs a dual-stream mixture-of-experts architecture on shared interleaved multimodal sequences, enabling joint context learning while decoupling the pathways for understanding and generation. We further introduce modality-aware rotary positional encoding to mitigate interference among heterogeneous visual tokens and boost cross-task alignment. During training, Lance adopts a staged multi-task training paradigm with capability-oriented objectives and adaptive data scheduling to strengthen both semantic comprehension and visual generation performance. Experimental results demonstrate that Lance substantially outperforms existing open-source unified models in image and video generation, while retaining strong multimodal understanding capabilities. The homepage is available at https://lance-project.github.io.