Search papers, labs, and topics across Lattice.
This paper introduces the WInning Noise Retrieval and Optimization (WINRO) framework, which enhances the alignment of text-to-motion synthesis by leveraging specific instances of Gaussian noise, termed "winning noise tickets." By mapping random noise to motion features and refining these selections through a KL-regularized objective, WINRO effectively reduces semantic drift and improves the fidelity of generated motions across various models without requiring retraining. The results demonstrate significant improvements in text-motion fidelity and temporal robustness, showcasing its versatility in applications like motion stylization and spatial constraints.
Winning noise tickets can dramatically enhance text-to-motion fidelity, bridging the gap between input semantics and generated motion without the need for retraining.
Diffusion-based text-to-motion models synthesize realistic human motions but often exhibit semantic drift from the input text. Motion is inherently temporal, especially in compositional and long-duration sequences that require semantic consistency across multiple action segments and smooth kinematic transitions throughout the trajectory. We posit that the initial noise is central to this consistency: within the Gaussian noise space, certain instances, i.e. winning noise tickets, carry latent structure that biases denoising toward particular motion semantics, even under null prompts. We propose WInning Noise Retrieval and Optimization (WINRO), a training-free, model-agnostic framework that improves text-motion alignment by selecting and refining such tickets before diffusion sampling. WINRO maps random noises to motion features generated under null prompts, retrieves the best-aligned noise for a given text, and refines it via a KL-regularized objective that reduces the residual semantic gap while preserving the Gaussian prior. An optional LoRA-based adapter amortizes this refinement into a single forward pass. WINRO consistently improves text-motion fidelity across different base models, MDM and MotionLCM, on HumanML3D without retraining, improves temporal robustness on the MTT benchmark, and generalizes to applications such as motion stylization and spatial constraint satisfaction.
