The Consistency Critic: Correcting Inconsistencies in Generated Images via Reference-Guided Attentive Alignment

2025-11-25

AI Summary

To address fine-grained detail inconsistency in generated images, this paper proposes ImageCritic—a reference-guided, multi-round local post-editing framework. Methodologically, it constructs a reference-degraded-target triplet dataset and introduces an attention alignment loss coupled with a detail encoder to precisely localize and rectify inconsistent regions. A vision-language-model-driven selection mechanism and explicit degradation strategy are incorporated, enabling progressive refinement within a proxy-based iterative optimization framework. Compared to existing approaches, ImageCritic significantly improves both detail consistency and visual quality across diverse customized generation scenarios. It is the first method to holistically integrate attention mechanisms, fine-grained detail encoding, and multi-round editing, establishing a scalable and interpretable paradigm for generative image post-processing.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Previous works have explored various customized generation tasks given a reference image, but they still face limitations in generating consistent fine-grained details. In this paper, our aim is to solve the inconsistency problem of generated images by applying a reference-guided post-editing approach and present our ImageCritic. We first construct a dataset of reference-degraded-target triplets obtained via VLM-based selection and explicit degradation, which effectively simulates the common inaccuracies or inconsistencies observed in existing generation models. Furthermore, building on a thorough examination of the model's attention mechanisms and intrinsic representations, we accordingly devise an attention alignment loss and a detail encoder to precisely rectify inconsistencies. ImageCritic can be integrated into an agent framework to automatically detect inconsistencies and correct them with multi-round and local editing in complex scenarios. Extensive experiments demonstrate that ImageCritic can effectively resolve detail-related issues in various customized generation scenarios, providing significant improvements over existing methods.

Problem

• Corrects inconsistencies in generated images
• Uses reference-guided post-editing approach
• Resolves fine-grained detail issues automatically

Innovation

• Reference-guided post-editing corrects image inconsistencies
• Attention alignment loss and detail encoder rectify fine details
• Multi-round local editing framework automatically detects and corrects