TAP: A Token-Adaptive Predictor Framework for Training-Free Diffusion Acceleration
This work addresses the problem of slow inference in diffusion models for machine learning practitioners and researchers, offering an incremental yet significant improvement in efficiency.
The authors tackled the problem of slow inference in diffusion models, achieving large speedups with little to no perceptual quality loss using their Token-Adaptive Predictor framework. The framework enables substantial improvements in the accuracy-efficiency frontier compared to fixed global predictors and caching-only baselines.
Diffusion models achieve strong generative performance but remain slow at inference due to the need for repeated full-model denoising passes. We present Token-Adaptive Predictor (TAP), a training-free, probe-driven framework that adaptively selects a predictor for each token at every sampling step. TAP uses a single full evaluation of the model's first layer as a low-cost probe to compute proxy losses for a compact family of candidate predictors (instantiated primarily with Taylor expansions of varying order and horizon), then assigns each token the predictor with the smallest proxy error. This per-token "probe-then-select" strategy exploits heterogeneous temporal dynamics, requires no additional training, and is compatible with various predictor designs. TAP incurs negligible overhead while enabling large speedups with little or no perceptual quality loss. Extensive experiments across multiple diffusion architectures and generation tasks show that TAP substantially improves the accuracy-efficiency frontier compared to fixed global predictors and caching-only baselines.