The Year in AI Papers
The year/Meta AI

Paper 2512.02014

TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models

Zhiheng Liu, Weiming Ren, Haozhe Liu, Zijian Zhou, Shoufa Chen, Haonan Qiu, Xiaoke Huang, Zhaochong An, Fanny Yang, Aditya Patel, Viktar Atliha, Tony Ng, et al.

Published
Dec 2025
Research lab
Meta AI
Citations
31
GitHub
94 stars

01 In brief

Summary

Tuna is a native unified multimodal model (UMM) that creates a unified continuous visual representation by cascading a VAE encoder with a representation encoder (SigLIP 2).

This design avoids the representation format mismatches of decoupled models, improving both understanding and generation.

The model uses an LLM decoder (Qwen2.5) for autoregressive text generation and a flow matching head for visual generation, trained in three stages.

Experiments show Tuna achieves state-of-the-art results on image/video understanding (e.g., 61.2% on MMStar), image generation (0.90 on GenEval), image editing, and video generation, outperforming prior UMMs like Show-o2.

Ablations confirm that unified representations outperform decoupled ones, stronger representation encoders yield better performance, and joint training on understanding and generation provides mutual benefits.

Tuna's unified representation shows higher alignment with generation model features than Show-o2's late-fusion approach, which remains biased toward semantics.

02 From the paper

Abstract

Unified multimodal models (UMMs) aim to jointly perform multimodal understanding and generation within a single framework. We present TUNA, a native UMM that builds a unified continuous visual representation by cascading a VAE encoder with a representation encoder. This unified representation space allows end-to-end processing of images and videos for both understanding and generation tasks. Compared to prior UMMs with decoupled representations, TUNA's unified visual space avoids representation format mismatches introduced by separate encoders, outperforming decoupled alternatives in both understanding and generation. Moreover, we observe that stronger pretrained representation encoders consistently yield better performance across all multimodal tasks, highlighting the importance of the representation encoder. Finally, in this unified setting, jointly training on both understanding and generation data allows the two tasks to benefit from each other rather than interfere. Our extensive experiments on multimodal understanding and generation benchmarks show that TUNA achieves state-of-the-art results in image and video understanding, image and video generation, and image editing, demonstrating the effectiveness and scalability of its unified representation design.