Free Lunch for Unified Multimodal Models: Enhancing Generation via Reflective Rectification with Inherent Understanding

2026-04-15

AI Summary

This work addresses the notable gap between comprehension and generation capabilities in unified multimodal models, where generative performance often fails to leverage the model’s strong inherent understanding. To bridge this gap, we propose UniRect-CoT, a novel framework that, for the first time, harnesses the model’s own comprehension as a self-supervised signal to iteratively refine its outputs through a Chain-of-Thought (CoT) mechanism—without requiring additional training. By integrating intrinsic visual reasoning from the diffusion denoising process with intermediate corrections aligned to instructions, UniRect-CoT achieves “zero-cost” performance gains. The framework seamlessly integrates into existing unified multimodal architectures and consistently enhances output quality across diverse and complex generation tasks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• 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
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Unified Multimodal Models (UMMs) aim to integrate visual understanding and generation within a single structure. However, these models exhibit a notable capability mismatch, where their understanding capability significantly outperforms their generation. This mismatch indicates that the model's rich internal knowledge, while effective for understanding tasks, remains underactivated during generation. To address this, we draw inspiration from the human ``Thinking-While-Drawing'' paradigm, where humans continuously reflect to activate their knowledge and rectify intermediate results. In this paper, we propose UniRect-CoT, a training-free unified rectification chain-of-thought framework. Our approach unlocks the ``free lunch'' hidden in the UMM's powerful inherent understanding to continuously reflect, activating its internal knowledge and rectifying intermediate results during generation.We regard the diffusion denoising process in UMMs as an intrinsic visual reasoning process and align the intermediate results with the target instruction understood by the model, serving as a self-supervisory signal to rectify UMM generation.Extensive experiments demonstrate that UniRect-CoT can be easily integrated into existing UMMs, significantly enhancing generation quality across diverse complex tasks.

Problem

• Unified Multimodal Models
• capability mismatch
• generation quality
• inherent understanding
• knowledge activation

Innovation

• Unified Multimodal Models
• Reflective Rectification
• Chain-of-Thought
• Diffusion Denoising
• Self-supervision