BRACE: Taming Sharp Irregularities via Barycentric Rational Forecasting for Fast Diffusion Transformers Inference

Researchers propose Barycentric Rational Forecasting with Chebyshev Enhancement (BRACE), a method that uses barycentric rational functions to forecast cached features in Diffusion Transformers (DiTs), improving inference speed while maintaining image and video generation quality. The method, detailed in arXiv:2608.07572v1, achieves state-of-the-art quality-efficiency trade-offs across various DiT architectures with negligible computational overhead.

arXiv:2608.07572v1 Announce Type: new Abstract: Diffusion Transformers DiTs have demonstrated exceptional performance in high-fidelity image and video generation. To alleviate their massive computational overhead, temporal feature caching has been proposed to bypass redundant computations. However, existing cache-then-forecast methods driven by derivative-based polynomials often cause severe quality degradation under high acceleration due to unstable long-step predictions. To address this bottleneck, we propose Barycentric Rational Forecasting with Chebyshev Enhancement BRACE . Motivated by the observation that DiT feature trajectories are globally smooth yet frequently exhibit sharp irregularities and local non-smoothness, BRACE shifts the paradigm from derivative-driven polynomial extrapolation to feature-driven rational forecasting. Specifically, it maintains a local sliding window to cache sparse historical features and leverages adapted Chebyshev weights to formulate a barycentric rational function, directly aggregating these raw features to ensure numerical stability. Extensive experiments demonstrate that BRACE achieves state-of-the-art quality-efficiency trade-offs across various DiT architectures with negligible computational overhead.