---
title: "DeCoRL: Decomposed Consistency Reinforcement Learning for Multi-Image Composition"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2610.39265v1"
permalink: "https://feed.astralor.com/items/itm_77cf99c247f090f84415596a"
published: "2026-10-29T03:18:03.000Z"
score: 54
channel: "tech"
tags: ["多模态","DeCoRL","Decomposed Consistency Reinforcement","Learning","Multi-Image Composition Multi-image","MICo","While","Multi-image"]
---

# DeCoRL: Decomposed Consistency Reinforcement Learning for Multi-Image Composition

## 速览

Multi-image composition (MICo) is a critical yet challenging capability for large-scale image editing models, where consistency degrades rapidly as the number of reference images increases. While supervised fine-tuning …

## 关键点

- Multi-image composition (MICo) is a critical yet challenging capability for large-scale image editing models, where consistency degrades rapidly as the number of reference images increases.
- While supervised fine-tuning (SFT) can equip models with basic MICo ability, it provides only implicit supervision on the composed image and often fails to learn reliable object-level correspondence across multiple refe…
- We propose DeCoRL, a reinforcement-learning (RL) framework that improves MICo consistency by decomposing holistic MICo consistency into object-level consistency.

## 全文

Multi-image composition (MICo) is a critical yet challenging capability for large-scale image editing models, where consistency degrades rapidly as the number of reference images increases. While supervised fine-tuning (SFT) can equip models with basic MICo ability, it provides only implicit supervision on the composed image and often fails to learn reliable object-level correspondence across multiple references. We propose DeCoRL, a reinforcement-learning (RL) framework that improves MICo consistency by decomposing holistic MICo consistency into object-level consistency. DeCoRL first establishes instruction-relevant object correspondences across reference images and localizes the corresponding regions in the generated image. It then uses a VLM-based reward model trained on single-object consistency annotations to produce fine-grained, object-wise feedback, which is aggregated into a multi-object consistency reward to guide RL post-training. We further introduce DeCo-Bench, a benchmark covering compositions from 2 to 6 reference images. Experiments on DeCo-Bench and the public MICo-Bench demonstrate consistent improvements in MICo consistency.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_77cf99c247f090f84415596a