Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Architecture Decoupling Is Not All You Need For Unified Multimodal Model

About

Unified multimodal models for image generation and understanding represent a significant step toward AGI and have attracted widespread attention from researchers. The main challenge of this task lies in the difficulty in establishing an optimal training paradigm due to inherent conflicting targets in understanding and generation tasks. To alleviate these conflicts and pursue higher performance, many researchers adopt varying degrees of model decoupling (e.g., Double image encoders, MOE/MOT architecture, or frozen MLLM). However, excessive model decoupling can lead to the loss of interleave generation ability, undermining the original intent of unified models. In this work, we aim to explore how to mitigate task conflicts without resorting to model decoupling. Firstly, we analyze why decoupling alleviates conflicts by studying the cross-modal attention behavior of models. We observe that model decoupling essentially drives models toward task-specific multimodal interaction patterns, as seen in Qwen-VL and HunyuanImage, and that the more thorough the decoupling, the more consistent the behavior becomes. Motivated by this observation, we propose Attention Interaction Alignment (AIA) loss, which explicitly learns Task-Specific multimodal interaction patterns during training. To demonstrate the generalizability of our AIA loss, we apply it to Emu3 and Janus-Pro during SFT and post-training stage respectively. Without bells and whistles, AIA not only refines cross-modal attention patterns, but also boosts both generation and understanding performance.

Dian Zheng, Manyuan Zhang, Hongyu Li, Kai Zou, Hongbo Liu, Ziyu Guo, Kaituo Feng, Yexin Liu, Ying Luo, Yan Feng, Peng Pei, Xunliang Cai, Hongsheng Li• 2025

Related benchmarks

TaskDatasetResultRank
Object Hallucination EvaluationPOPE--
935
Multimodal UnderstandingMM-Vet
MM-Vet Score49.8
418
Multimodal UnderstandingMMBench--
367
Multimodal UnderstandingMMMU (test)
MMMU Score42.1
86
Multimodal Visual PerceptionMMVP
Accuracy48
44
Image GenerationGenEval overall
GenEval Overall Score81
30
Image GenerationDPG
DPG Score84.49
20
Multimodal EvaluationMME-P
MME-P Score1.66e+3
14
Showing 8 of 8 rows

Other info

GitHub

Follow for update