CoLoGen: Progressive Learning of Concept`-`Localization Duality for Unified Image Generation

2026-02-25

AI Summary

Unified conditional image generation struggles to reconcile conflicting requirements between heterogeneous representations—such as conceptual understanding and spatial localization—within a single shared latent space. To address this challenge, this work proposes CoLoGen, a novel framework that introduces concept–localization duality modeling and a Progressive Representation Weaving (PRW) mechanism. Through curriculum-based, stage-wise learning, CoLoGen dynamically routes features to specialized expert modules, enabling joint optimization of both capabilities. Built upon a diffusion model architecture, the proposed method achieves state-of-the-art or superior performance across diverse tasks including image editing, controllable generation, and customized synthesis, thereby demonstrating its effectiveness and generalizability.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Unified conditional image generation remains difficult because different tasks depend on fundamentally different internal representations. Some require conceptual understanding for semantic synthesis, while others rely on localization cues for spatial precision. Forcing these heterogeneous tasks to share a single representation leads to concept`-`localization representational conflict. To address this issue, we propose CoLoGen, a unified diffusion framework that progressively learns and reconciles this concept`-`localization duality. CoLoGen uses a staged curriculum that first builds core conceptual and localization abilities, then adapts them to diverse visual conditions, and finally refines their synergy for complex instruction`-`driven tasks. Central to this process is the Progressive Representation Weaving (PRW) module, which dynamically routes features to specialized experts and stably integrates their outputs across stages. Experiments on editing, controllable generation, and customized generation show that CoLoGen achieves competitive or superior performance, offering a principled representational perspective for unified image generation.

Problem

• unified image generation
• concept-localization duality
• representational conflict
• conditional image generation
• diffusion models

Innovation

• concept-localization duality
• progressive representation weaving
• unified image generation
• diffusion framework
• staged curriculum learning