基于稀疏 Hopfield 检索的扩散模型方向性注意力引导（DAG）方法
Directional Attention Guidance for Diffusion Models via Sparse Hopfield Retrieval
AI 导读
研究者提出 Directional Attention Guidance（DAG），一种免训练、单次推理的注意力引导规则，通过 Modern Hopfield 动力学分析稀疏-密集注意力差异的方向性质，仅放大与检索方向对齐的分量。
正文 · 原文
Abstract:Classifier-Free Guidance (CFG) improves sample quality in diffusion models, but its dual-pass inference and null-condition training limit its use in few-step regimes. Attention-space guidance is a single-pass alternative, yet why sparse-versus-dense attention guidance works remains elusive. We analyze attention extrapolation through Modern Hopfield dynamics and establish two directional properties of the sparse-dense discrepancy under shared conditioning that identify it as an acceleration direction toward the dominant stored pattern. Building on this, we propose Directional Attention Guidance (DAG), a training-free, single-pass rule that amplifies only the component of the discrepancy aligned with the retrieval direction; stability follows from a weak contraction property, and an analysis inside real cross-attention layers shows that this alignment is observed throughout sampling. DAG generalizes across UNet and MMDiT backbones and multi-step and few-step regimes, improving text alignment and human-preference scores on every backbone tested, including FLUX.1, FLUX.2, and Qwen-Image, at a small added cost per image over the base sampler.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2603.02531 [cs.LG] |
| (or arXiv:2603.02531v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2603.02531 arXiv-issued DOI via DataCite |
Submission history
From: Kwanyoung Kim Professor [view email]
[v1]
Tue, 3 Mar 2026 02:44:21 UTC (8,504 KB)
[v2]
Sun, 17 May 2026 04:50:49 UTC (28,735 KB)
[v3]
Mon, 26 Oct 2026 13:40:23 UTC (8,957 KB)
来源：arXiv cs.AI · arxiv.org