aiqbench analysis
This research addresses the computational bottleneck in Diffusion Transformers by replacing unstable polynomial extrapolation with barycentric rational forecasting to maintain numerical stability during high-speed inference. By leveraging Chebyshev weights and local feature caching, the proposed BRACE method achieves state-of-the-art quality-efficiency trade-offs without adding significant overhead. This advancement is critical for deploying high-fidelity generative models in latency-sensitive applications where traditional acceleration techniques degrade output quality.
BRACE: Taming Sharp Irregularities via Barycentric Rational Forecasting for Fast Diffusion Transformers Inference
arXiv:2608.07572v2 Announce Type: replace-cross 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.
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.
Accepted at ACM MM 2026. Project page: this https URL
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2608.07572 [cs.CV]
(or arXiv:2608.07572v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2608.07572 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 251 words
