Jano: Adaptive Diffusion Generation with Early-stage Convergence Awareness

2026-02-28

AI Summary

Diffusion models suffer from computational inefficiency during generation, particularly in Diffusion Transformers that rely on full attention mechanisms. Existing acceleration methods often employ content-agnostic uniform strategies, which fail to account for the heterogeneous convergence behavior across different image regions. To address this limitation, this work proposes Jano—an adaptive, training-free generation framework that leverages the spatially varying convergence characteristics inherent in the diffusion process. By dynamically identifying region-wise complexity early and adaptively scheduling token computation at runtime, Jano enables content-aware allocation of computational resources. Experiments demonstrate that Jano achieves an average speedup of 2.0× (up to 2.4×) over standard pipelines on mainstream models while preserving generation quality.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Web Mining and Content Analysis: Content-based information diffusion
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Diffusion models have achieved remarkable success in generative AI, yet their computational efficiency remains a significant challenge, particularly for Diffusion Transformers (DiTs) requiring intensive full-attention computation. While existing acceleration approaches focus on content-agnostic uniform optimization strategies, we observe that different regions in generated content exhibit heterogeneous convergence patterns during the denoising process. We present Jano, a training-free framework that leverages this insight for efficient region-aware generation. Jano introduces an early-stage complexity recognition algorithm that accurately identifies regional convergence requirements within initial denoising steps, coupled with an adaptive token scheduling runtime that optimizes computational resource allocation. Through comprehensive evaluation on state-of-the-art models, Jano achieves substantial acceleration (average 2.0 times speedup, up to 2.4 times) while preserving generation quality. Our work challenges conventional uniform processing assumptions and provides a practical solution for accelerating large-scale content generation. The source code of our implementation is available at https://github.com/chen-yy20/Jano.

Problem

• diffusion models
• computational efficiency
• convergence heterogeneity
• Diffusion Transformers
• content generation

Innovation

• adaptive diffusion
• early-stage convergence awareness
• region-aware generation
• token scheduling
• training-free acceleration