Test-Time Iterative Error Correction for Efficient Diffusion Models

2025-11-09

AI Summary

Efficient diffusion models suffer from exponential error accumulation across denoising steps when deployed on resource-constrained devices, due to approximation-based acceleration techniques—degrading generation quality severely and resisting post-hoc correction. This work is the first to characterize this error propagation dynamics. We propose a test-time iterative correction method that requires no retraining or architectural modification: during inference, it dynamically analyzes and rectifies outputs at each timestep, suppressing error growth from exponential to linear. Our approach is fully compatible with mainstream efficiency optimizations—including step pruning and knowledge distillation—as well as diverse model architectures. Extensive experiments across multiple datasets and models demonstrate consistent improvements: FID scores decrease by 12.3%–28.7% while preserving inference efficiency. The method enables flexible trade-offs between accuracy and computational cost without sacrificing deployment practicality.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

With the growing demand for high-quality image generation on resource-constrained devices, efficient diffusion models have received increasing attention. However, such models suffer from approximation errors introduced by efficiency techniques, which significantly degrade generation quality. Once deployed, these errors are difficult to correct, as modifying the model is typically infeasible in deployment environments. Through an analysis of error propagation across diffusion timesteps, we reveal that these approximation errors can accumulate exponentially, severely impairing output quality. Motivated by this insight, we propose Iterative Error Correction (IEC), a novel test-time method that mitigates inference-time errors by iteratively refining the model's output. IEC is theoretically proven to reduce error propagation from exponential to linear growth, without requiring any retraining or architectural changes. IEC can seamlessly integrate into the inference process of existing diffusion models, enabling a flexible trade-off between performance and efficiency. Extensive experiments show that IEC consistently improves generation quality across various datasets, efficiency techniques, and model architectures, establishing it as a practical and generalizable solution for test-time enhancement of efficient diffusion models.

Problem

• Correcting approximation errors in efficient diffusion models during inference
• Mitigating exponential error accumulation across diffusion timesteps
• Enhancing image quality without retraining or architectural modifications

Innovation

• Iterative Error Correction reduces exponential error growth
• Method refines outputs during inference without retraining
• Seamlessly integrates with existing diffusion model architectures