Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models

2026-07-05

AI Summary

Existing diffusion model inference acceleration methods overly rely on intermediate verification, neglecting the computational cost of generation itself and thereby distorting efficiency evaluations. This work proposes Flash-BoN, a framework that unifies for the first time three acceleration strategies—timestep truncation, network layer skipping, and activation surrogation—and introduces a low-cost draft candidate pool coupled with a multi-stage verification mechanism. Under a fixed wall-clock time budget, Flash-BoN prioritizes breadth-wise exploration over frequent validation. The method consistently outperforms existing approaches across three benchmarks and three model scales, achieving an 8% AUC improvement in large models. When combined with orthogonal techniques such as prompt optimization, it yields a 16% AUC gain, while also revealing the critical role of candidate diversity in generation quality and convergence of reinforcement learning-based post-training.

Abstract

Inference-time scaling for text-to-image generation has progressed from simple Best-of-$N$ (BoN) sampling to guided search methods that verify and steer candidate trajectories at intermediate denoising steps. These approaches focus on when and how often to verify during denoising but largely treat the cost of generation itself as fixed. Moreover, the standard practice of comparing methods by number of function evaluations (NFEs) counts only denoising forward passes and ignores verifier overhead, which can distort efficiency rankings. We show that under wall-clock evaluation, simple BoN already matches or outperforms several guided search techniques, suggesting that compute is better spent on broader exploration than on repeated intermediate verification. This motivates Flash-BoN, which generates a large pool of inexpensive draft candidates by combining three complementary acceleration knobs: timestep truncation, layer skipping, and activation proxies into a single configuration optimized once per model. An efficient multi-stage verification procedure then identifies the most promising draft, which is refined at full quality. Across three benchmarks and three model scales, Flash-BoN consistently outperforms all baselines under fixed wall-clock budgets, with gains that grow at larger model scales (+8% AUC). We further show that our strategy combines well and improves existing orthogonal techniques such as reflection-based prompt optimization (+16% AUC). The gains correlate with increased candidate diversity, which also enables draft-guided selection to accelerate RL post-training convergence.

Problem

• inference-time scaling
• diffusion models
• text-to-image generation
• function evaluations
• verification overhead

Innovation

• inference-time scaling
• diffusion models
• draft generation
• efficient verification
• wall-clock efficiency