SAGE: Semantic-Aware Shared Sampling for Efficient Diffusion

2025-09-19

AI Summary

Diffusion models suffer from low sampling efficiency due to their reliance on dozens of iterative denoising steps. To address this, we propose a semantic-aware shared sampling framework—the first to enable semantically similar input queries to jointly traverse a common denoising path during early sampling stages, thereby departing from the conventional “per-query independent inference” paradigm. To realize this mechanism, we introduce a customized training strategy that jointly optimizes shared-path representation learning and semantic similarity modeling. Our approach preserves generation fidelity while substantially improving both sampling efficiency and output diversity. Experiments demonstrate a 25.5% reduction in sampling steps, a 5.0-point improvement in FID, a 5.4-point gain in CLIP Score, and a 160% increase in diversity metrics. This work establishes a novel paradigm for efficient diffusion sampling.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders

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: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Diffusion models manifest evident benefits across diverse domains, yet their high sampling cost, requiring dozens of sequential model evaluations, remains a major limitation. Prior efforts mainly accelerate sampling via optimized solvers or distillation, which treat each query independently. In contrast, we reduce total number of steps by sharing early-stage sampling across semantically similar queries. To enable such efficiency gains without sacrificing quality, we propose SAGE, a semantic-aware shared sampling framework that integrates a shared sampling scheme for efficiency and a tailored training strategy for quality preservation. Extensive experiments show that SAGE reduces sampling cost by 25.5%, while improving generation quality with 5.0% lower FID, 5.4% higher CLIP, and 160% higher diversity over baselines.

Problem

• Reducing high sampling cost of diffusion models
• Sharing early-stage sampling across similar queries
• Maintaining generation quality while improving efficiency

Innovation

• Semantic-aware shared sampling framework
• Tailored training strategy for quality
• Reduces sampling cost by 25.5%