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

Hugging Face Daily Papers Papers
text-to-image diffusion-models inference-time-scaling best-of-n draft-candidates efficient-generation

Summary

Flash-BoN improves text-to-image generation efficiency by generating cheap draft candidates via timestep truncation, layer skipping, and activation proxies, then using multi-stage verification to select the best draft for full refinement, outperforming baselines under fixed wall-clock budgets.

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.
Original Article
View Cached Full Text

Cached at: 07/10/26, 06:17 AM

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

Source: https://huggingface.co/papers/2607.04461

Abstract

Flash-BoN improves text-to-image generation efficiency by using inexpensive draft candidates generated through timestep truncation, layer skipping, and activation proxies, followed by multi-stage verification that outperforms existing methods under fixed wall-clock budgets.

Inference-time scaling for text-to-image generation has progressed from simpleBest-of-N(BoN) sampling toguided search methodsthat verify and steer candidate trajectories at intermediatedenoising 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 offunction evaluations(NFEs) counts only denoising forward passes and ignores verifier overhead, which can distort efficiency rankings. We show that underwall-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 motivatesFlash-BoN, which generates a large pool of inexpensive draft candidates by combining three complementary acceleration knobs:timestep truncation,layer skipping, andactivation proxiesinto a single configuration optimized once per model. An efficientmulti-stage verificationprocedure then identifies the most promising draft, which is refined at full quality. Across three benchmarks and three model scales,Flash-BoNconsistently 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 increasedcandidate diversity, which also enables draft-guided selection to accelerateRL post-training convergence.

View arXiv pageView PDFProject pageGitHub3Add to collection

Get this paper in your agent:

hf papers read 2607\.04461

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2607.04461 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2607.04461 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.04461 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

DFlash: Block Diffusion for Flash Speculative Decoding

Papers with Code Trending

DFlash is a new speculative decoding framework that uses a lightweight block diffusion model for parallel token drafting, achieving over 6x acceleration compared to autoregressive methods. It significantly outperforms existing state-of-the-art methods like EAGLE-3 while maintaining high output quality.

Fast Byte Latent Transformer

Hugging Face Daily Papers

This paper introduces BLT Diffusion and speculative decoding techniques for byte-level language models to significantly reduce generation latency and memory bandwidth costs while maintaining quality.