Hallucination Early Detection in Diffusion Models

2026-04-22

AI Summary

This work addresses the tendency of diffusion models to omit specified entities when generating multi-object images. To mitigate this issue, the authors propose the HEaD+ framework, which integrates cross-attention maps, textual prompts, and a predicted final image during early diffusion steps. By verifying object presence and spatial relationships, HEaD+ enables early hallucination detection and implements a generation gating mechanism that dynamically decides whether to restart the sampling process. Notably, this approach is the first to leverage the predicted image as an intermediate criterion for such decisions. Trained on a newly introduced InsideGen dataset, HEaD+ significantly improves both generation completeness and efficiency: in four-object tasks, it increases the success rate of fully complete generations by 6–8% and reduces the time required to achieve target images by up to 32%.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Search and Optimization: Heuristic Search

Application Category

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

Abstract

Text-to-Image generation has seen significant advancements in output realism with the advent of diffusion models. However, diffusion models encounter difficulties when tasked with generating multiple objects, frequently resulting in hallucinations where certain entities are omitted. While existing solutions typically focus on optimizing latent representations within diffusion models, the relevance of the initial generation seed is typically underestimated. While using various seeds in multiple iterations can improve results, this method also significantly increases time and energy costs. To address this challenge, we introduce HEaD+ (Hallucination Early Detection +), a novel approach designed to identify incorrect generations early in the diffusion process. The HEaD+ framework integrates cross-attention maps and textual information with a novel input, the Predicted Final Image. The objective is to assess whether to proceed with the current generation or restart it with a different seed, thereby exploring multiple-generation seeds while conserving time. HEaD+ is trained on the newly created InsideGen dataset of 45,000 generated images, each containing prompts with up to seven objects. Our findings demonstrate a 6-8% increase in the likelihood of achieving a complete generation (i.e., an image accurately representing all specified subjects) with four objects when applying HEaD+ alongside existing models. Additionally, HEaD+ reduces generation times by up to 32% when aiming for a complete image, enhancing the efficiency of generating complete and accurate object representations relative to leading models. Moreover, we propose an integrated localization module that predicts object centroid positions and verifies pairwise spatial relations (if requested by the users) at an intermediate timestep, gating generation together with object presence to further improve relation-consistent outcomes.

Problem

• hallucination
• diffusion models
• text-to-image generation
• object omission
• generation seed

Innovation

• Hallucination Detection
• Diffusion Models
• Early Stopping
• Cross-Attention Analysis
• Object Localization