# BeCARE: Budgeted Cache Refresh for Diffusion Transformer Acceleration

arXiv: https://arxiv.org/abs/2605.27075 · HTML: https://mlsanity.com/paper/2605.27075/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 20 of its 27 references that are in the map.
- First posted on arXiv 26 May 2026.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)

## Builds on (most cited first, 20 of 20)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps](https://mlsanity.com/paper/2206.00927/) (2022)
- [Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model](https://mlsanity.com/paper/2411.19108/) (2024)
- [Accelerating Diffusion Transformers with Token-wise Feature Caching](https://mlsanity.com/paper/2410.05317/) (2024)
- [From Reusing to Forecasting: Accelerating Diffusion Models with TaylorSeers](https://mlsanity.com/paper/2503.06923/) (2025)
- [SpeCa: Accelerating Diffusion Transformers with Speculative Feature Caching](https://mlsanity.com/paper/2509.11628/) (2025)
- [Forecast then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers](https://mlsanity.com/paper/2508.16211/) (2025)
- [FreqCa: Accelerating Diffusion Models via Frequency-Aware Caching](https://mlsanity.com/paper/2510.08669/) (2025)
- [Compute Only 16 Tokens in One Timestep: Accelerating Diffusion Transformers with Cluster-Driven Feature Caching](https://mlsanity.com/paper/2509.10312/) (2025)
- [Accelerating Diffusion Transformer via Gradient-Optimized Cache](https://mlsanity.com/paper/2503.05156/) (2025)
- [Accelerating Diffusion Transformer via Error-Optimized Cache](https://mlsanity.com/paper/2501.19243/) (2025)
- [Relational Feature Caching for Accelerating Diffusion Transformers](https://mlsanity.com/paper/2602.19506/) (2026)
- [Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Efficient Diffusion Transformers](https://mlsanity.com/paper/2601.07396/) (2026)
- [TAP: A Token-Adaptive Predictor Framework for Training-Free Diffusion Acceleration](https://mlsanity.com/paper/2603.03792/) (2026)
- [Predict to Skip: Linear Multistep Feature Forecasting for Efficient Diffusion Transformers](https://mlsanity.com/paper/2602.18093/) (2026)
- [Accelerating Controllable Generation via Hybrid-grained Cache](https://mlsanity.com/paper/2511.11031/) (2025)

## Built on by (most cited first, 1 of 1)

- [OmniCache: Multidimensional Hierarchical Feature Caching For Diffusion Models](https://mlsanity.com/paper/2607.23844/) (2026)

## Abstract (arXiv)

Training-free feature caching accelerates diffusion transformer (DiT) inference by reusing or forecasting intermediate features. However, fixed schedules make compute predictable but ignore prompt- and timestep-dependent risk, whereas hand-tuned error thresholds adapt locally but leave realized compute difficult to control. To address these limitations, we present Budgeted Cache Refresh (BeCARE), a training-free framework that, given a user-specified cap on full computations, adaptively determines when to refresh the cache during accelerated inference. Specifically, we first derive an error-amplification profile from the sampler's noise schedule to characterize the varying impact of approximation errors across denoising timesteps. We then combine this profile with prompt-specific extrapolation residuals and cache age to form an accumulated risk score, triggering a refresh when continued caching becomes harmful. Meanwhile, we use the same profile to construct an analytic spending reference and adjust the refresh threshold through feedback, thereby distributing the limited refresh budget over the sampling trajectory. Finally, structural safeguards (a fixed warmup, a budget-derived late-stage reserve, and a maximum cache-reuse length) prevent unreliable extrapolation and premature budget exhaustion. These designs enable prompt-adaptive refresh placement and allow a single calibration to transfer across budget tiers without per-tier tuning. Experiments on FLUX.1-dev with 200 prompts and measured FLOPs show that our method consistently outperforms representative training-free caching baselines across speedups from 3x to 6x. At the main operating point (3.3x acceleration), it improves paired PSNR by 2.7 dB over the strongest baseline using no more compute. On SD3.5 Large, it also substantially improves the same Taylor cache engine over fixed-interval scheduling at matched compute.