---
title: "ChebBooster: A Training-Free Approach for Efficient Diffusion Transformer Inference via Chebyshev-Inspired Extrapolation"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.23429v1"
permalink: "https://feed.astralor.com/items/itm_8593690f937a61844df53d2e"
published: "2026-08-24T16:06:11.000Z"
score: 75
channel: "ai"
tags: ["ChebBooster","Chebyshev Polynomial","Image Generation","Diffusion Transformer","Training-Free Acceleration"]
---

# ChebBooster: A Training-Free Approach for Efficient Diffusion Transformer Inference via Chebyshev-Inspired Extrapolation

## AI commentary

> We think ChebBooster's highlight is not just the extrapolation trick, but turning numerical stability into an engineering problem; offline weight precomputation adds almost zero online overhead. For engineers running DiT inference, this is a directly testable acceleration option, and it’s worth watching whether it can stack with distillation or quantization.

## At a glance

Diffusion Transformers (DiTs) excel at high-fidelity image generation but suffer from high inference costs due to full model execution at every timestep. Existing cache-based acceleration suffers from low accuracy over long intervals, while Taylor-series extrapolation often faces instability from Runge oscillations. ChebBooster proposes a training-free extrapolation framework grounded in Chebyshev polynomial theory, using barycentric interpolation for numerical stability and low overhead. It decouples extrapolation into offline weight precomputation and a lightweight online stage. Across DiT-XL/2, PixArt-Σ, and FLUX.1-dev, ChebBooster achieves up to 3.68× latency speedup and 5.12× FLOPs reduction, consistently outperforming existing training-free baselines across tasks and resolutions.

## Key points

- Proposes a training-free acceleration framework using Chebyshev polynomial extrapolation, eliminating training overhead
- Employs barycentric formulation for high numerical stability with minimal online cost
- Achieves up to 3.68× latency speedup and 5.12× FLOPs reduction on DiT-XL/2, PixArt-Σ, and FLUX.1-dev
- Decouples extrapolation into offline weight precomputation and a lightweight online stage for practical deployment

## Full text

Diffusion Transformers (DiTs) have shown strong performance in high-fidelity image generation, but their sampling process remains computationally intensive due to full model execution at every timestep. While cache-based acceleration has been explored to mitigate inference cost, naive reuse schemes suffer from low accuracy over long intervals, and Taylor-series-based extrapolation methods often face instability caused by Runge oscillations. In this paper, we propose ChebBooster, a training-free extrapolation framework based on Chebyshev polynomial theory that achieves stable and efficient acceleration for DiTs. Specifically, we adopt the Barycentric formulation to evaluate Chebyshev approximants with high numerical stability and minimal overhead, and further decouple the extrapolation into an offline weight precomputation phase and a lightweight online application stage. Extensive experiments across three representative DiT-based models, including DiT-XL/2, PixArt-$Σ$, and FLUX.1-dev, demonstrate that ChebBooster achieves consistent improvements in visual quality and inference efficiency, reaching up to $3.68\times$ latency speedup and $5.12\times$ FLOPs reduction, outperforming existing training-free baselines under diverse generation tasks and resolutions.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_8593690f937a61844df53d2e