Beyond Structural Unification: A Perspective and Roadmap on Unified Multimodal Models

Sep 30, 2026·
Yinyi Luo
,
Hayes Bai
Yiqiao Jin
Yiqiao Jin
,
Wenyue Hua
,
Srijan Kumar
,
Marios Savvides
,
Qingsong Wen
,
Jindong Wang
· 1 min read
Abstract
This position paper develops a roadmap for functional unification in unified multimodal models. It organizes the research agenda around five layers: modality-aware input grounding, coordinated representations for understanding and generation, stateful multimodal reasoning, joint optimization, and system-level evaluation. These layers connect architectural choices to the preservation and combination of model capabilities.
Type
Publication
NeurIPS 2026 Workshop on Linguistic Principles for Foundation Models

Overview

This position paper develops a roadmap for functional unification in unified multimodal models. It organizes the research agenda around five layers: modality-aware input grounding, coordinated representations for understanding and generation, stateful multimodal reasoning, joint optimization, and system-level evaluation. These layers connect architectural choices to the preservation and combination of model capabilities.

Multimodal Models Position Paper Unified Models Evaluation
Yiqiao Jin
Authors
Yiqiao Jin
Final-year Ph.D. Candidate in Computer Science
My research sits at the intersection of multimodal foundation models, intelligent agents, and spatial/physical intelligence. I build general-purpose models that perceive, reason, learn, and act in interactive virtual and physical environments.