LatentUMM: Dual Latent Alignment for Unified Multimodal Models

TL;DR

LatentUMM introduces a dual latent alignment framework to explicitly align transformations in shared latent spaces, enhancing cross-modal consistency in unified multimodal models.

Contribution

It proposes a novel dual latent alignment approach with latent dynamics stabilization to improve semantic consistency in multimodal models.

Findings

Improves cross-modal semantic consistency across architectures

Enhances robustness via stochastic latent rollouts

Achieves better alignment between generation and re-encoding processes

Abstract

Unified multimodal models (UMMs) achieve strong performance in both understanding and generation by learning a shared latent space, yet they often exhibit functional inconsistency between these two capabilities. We observe that this issue does not stem from a lack of shared representations, but from the absence of explicit alignment between the transformations that map into and out of the latent space. As a result, generation and re-encoding can follow inconsistent trajectories, leading to semantic drift under modality transitions. In this work, we propose LatentUMM, a framework that constructs an enhanced shared latent space to explicitly align these transformations and improve cross-modal consistency. LatentUMM consists of two stages. First, dual latent alignment enforces consistency at both the modality and capacity levels: cross-modal alignment uses a stronger embedding model to impose structured cross-modal semantics, while dual capacity alignment enforces bidirectional consistency under generation and re-encoding. Second, latent dynamics stabilization improves robustness via stochastic latent rollouts and preference optimization, favoring trajectories that better preserve semantic consistency. Experiments show that LatentUMM consistently improves multimodal consistency across diverse architectures. Code is available at: https://github.com/AIFrontierLab/TorchUMM/tree/main/src/umm/post_training/LatentUMM.