MUSE: Unlocking Timestep as Native Task Steering for One-Step Dense Prediction
This work provides a simple, parameter-free method for multi-task dense prediction, offering an efficient path toward generalist vision models by repurposing existing diffusion infrastructure.
The paper reveals that the fixed sinusoidal timestep embedding in one-step diffusion models can serve as a native task steering signal, enabling parameter-free multi-task dense prediction. MUSE achieves competitive performance on depth and normal estimation across 10 datasets without additional parameters.
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.