用时间步作为任务控制信号，实现无需额外参数的多任务图像预测。
MUSE: Unlocking Timestep as Native Task Steering for One-Step Dense Prediction
- 利用扩散模型中固定的时间步嵌入作为任务引导信号，无需新增参数。
- 在10个数据集上达到与主流方法相当的深度和法向估计性能。
- 适用于U-Net和DiT架构，适合追求高效多任务模型的研究者。
单目密集预测近年来通过复用预训练扩散模型取得了显著进展，这为更高效的多任务学习范式开辟了前景。然而，现有方法常引入参数量大的适配器、专家或可学习任务标记，导致计算冗余。本文揭示了一步扩散模型中的内在机制：原始的固定正弦时间步嵌入可被重新用作内生的任务控制信号。基于此发现，我们提出多任务统一估计方法MUSE，一种无参数、单模型的多任务密集预测方案。通过流形解耦理论解释该机制，离散且固定的时序值能确定性地引导生成过程进入潜空间中解耦的任务特定流形。在10个数据集上的广泛实验表明，MUSE在单目深度与法向估计任务上均达到高度竞争力的表现，且在U-Net与DiT架构间具有良好的泛化能力。本工作通过解锁现有生成基础设施的潜在能力，为通用视觉模型提供了一条简洁高效的路径。
原文摘要 · Abstract (English)
Monocular dense prediction has recently seen remarkable success by repurposing pre-trained diffusion models. This opens a promising yet challenging avenue for more efficient multi-task learning paradigm. However, existing multi-task diffusion methods often introduce parameter-heavy adapters, experts, or learnable task tokens, leading to computational redundancy. In this paper, we reveal an inherent mechanism within one-step diffusion models: the native, fixed sinusoidal timestep embedding can be repurposed as an endogenous task steering signal. Based on this discovery, we propose Multi-task Unified eStimation via timestep Embedding (MUSE), a parameter-free, single-model multi-tasking approach for dense prediction. We interpret this mechanism via Manifold Decoupling, where discrete, fixed timestep values deterministically steer the generation process towards decoupled, task-specific manifolds in the latent space. Extensive experiments across 10 datasets demonstrate that MUSE achieves highly competitive performance on both monocular depth and normal estimation, and its efficacy generalizes across U-Net and DiT architectures. Our work offers a concise and efficient path toward generalist vision models by simply unlocking the latent potential of existing generation infrastructure.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。