AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

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

Ruchit Rawal, Reza Shirkavand, Sayak Paul, Yuxin Wen, Heng Huang, Yizheng Chen, Tom Goldstein, Gowthami Somepalli

ECCV 2026 · Image generation › Fast diffusion sampling · approach: Training-free diffusion acceleration · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use optimized draft pool (Flash-BoN) instead of plain BoN under wall-clock budgetBest-of-N · GenAI/GenEval/UniGenBench, 3 modelsAUC/Time: e.g. 0.60 -> 0.68 (Wan 14B GenAI-Bench); 0.65 -> 0.66 (Wan 1.3B)Improved
use guided search (BFS/DFS/ZOS) instead of BoN under wall-clock budgetBoN · GenAI-Bench, Wan2.1 14BAUC/Time: BoN 0.60 vs BFS 0.43, DFS 0.50, ZOS 0.57Hurt
add layer skipping to timestep truncation drafts (ablation)Flash-BoN drafts · GenAI-Bench, Wan2.1 1.3BAUC: 0.623 -> 0.646Improved
add activation proxies to TS+LS (ablation)Flash-BoN drafts · GenAI-Bench, Wan2.1 1.3BAUC: 0.646 -> 0.652Improved
add multi-stage pairwise verification to three knobs (ablation)Flash-BoN · GenAI-Bench, Wan2.1 1.3BAUC: 0.652 -> 0.666Improved
use random draft configuration instead of optimized configuration (ablation)Flash-BoN · GenAI-Bench, Wan2.1 1.3BAUC: 0.666 -> 0.551Hurt
use FLUX.1-dev draft config on Wan2.1 14B instead of native calibration (ablation)Flash-BoN · GenAI-Bench, Wan2.1 14BAUC: 0.68 -> 0.61Hurt
layer Flash drafts onto Reflection-TuningReflection-Tuning · GenAI-BenchAUC: 0.46 -> 0.62Improved
layer Flash drafts onto BFSBFS · GenAI-BenchAUC: 0.49 -> 0.55Improved
replace random rollouts with draft-guided selection (16 drafts, top 6 + bottom 2)Flow-GRPO · Wan2.1 1.3B RL post-trainingreward: 0.692 -> 0.699; matches baseline final at step 60Improved
use larger 72B verifier instead of 7B (ablation)Flash-BoN · GenAI-Bench, Wan2.1 1.3Bscore under wall-clockHurt
use 3B verifier instead of 7B (ablation)Flash-BoN · GenAI-Bench, Wan2.1 1.3Bscore under wall-clockHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy