ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE

• URL: http://arxiv.org/abs/2606.06060v1
• Date: Thu, 04 Jun 2026 11:59:56 GMT
• Title: ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE
• Abstract summary: ReCache learns the recomputation schedule that maximizes generation quality.<n>It trains via policy gradients, sidestepping backpagation through full diffusion inference.<n>It consistently outperforms scheduling baselines.
• Score: 40.4732814295941
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Modern diffusion models generate high-quality images and videos, but their iterative denoising process makes inference expensive. Feature caching accelerates sampling by reusing or predicting intermediate activations across neighboring denoising steps, exploiting the redundancy of computations along the reverse trajectory. In this work, we focus on the caching schedule: selecting which denoising steps should be fully recomputed. Existing schedules are either fixed (e.g. uniform) or chosen adaptively from per-step error heuristics; in both cases, the actual compute cost is a side-effect of hand-tuned thresholds rather than a quantity the user can specify. We propose ReCache, which inverts this: given a target budget k, it learns the recomputation schedule that maximizes generation quality, turning compute into a directly controllable input. ReCache trains via policy gradients, sidestepping backpropagation through full diffusion inference, and uses no labelled data. Generations from uncached inference serve as matching targets, paired with a reward for generation quality. ReCache is compatible with any caching mechanism, including feature reuse and feature forecasting; for each mechanism, a single trained policy adapts across computational budgets at inference time. ReCache consistently outperforms scheduling baselines: under a $\times5.04$ FLOPs reduction on FLUX, it reduces LPIPS by 31% (from 0.456 to 0.316) compared to DiCache; on Wan 2.1 at a $\sim \times2.6$ speedup, it drops LPIPS by 65% (from 0.480 to 0.169) and boosts the VBench score by 7% (5.6 points, from 70.4 to 76.0) over uniform HiCache. Code is available at https://github.com/thecrazymage/ReCache.

Related papers

• EpaCache: Error-Propagation-Aware Caching for Accelerating Diffusion-Based Visual Generation [30.438361624564553]
  Caching-based methods reduce inference latency by reusing intermediate computations across adjacent timesteps.<n>EpaCache is a training-free caching policy that adaptively allocates the reuse budget on timesteps with lower downstream impact.<n>EpaCache consistently improves the latency--fidelity trade-off over existing caching methods.
  arXiv  Detail & Related papers  (2026-08-29T13:38:11Z)
• OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Models [11.053622780931306]
  OTCache is a training-free framework for accelerating diffusion sampling via caching schedule prediction.<n>We show that OTCache consistently improves generation fidelity over state-of-the-art caching baselines.
  arXiv  Detail & Related papers  (2026-06-30T01:46:54Z)
• Budget-Constrained Step-Level Diffusion Caching [9.784100090561045]
  Existing methods make per-step cache decisions using threshold-baseds, without directly optimizing for final output quality.<n>We propose BudCache, which fixes the compute budget in advance and search for the cache policy that best preserves the final output.<n>BudCache achieves better generation quality than deterministic caching baselines under the same inference budgets.
  arXiv  Detail & Related papers  (2026-06-11T15:45:05Z)
• Motion-Aware Caching for Efficient Autoregressive Video Generation [73.27762884730272]
  MotionCache exploits inter-frame differences as a lightweight proxy for pixel-level motion characteristics.<n>Experiments on state-of-the-art models like SkyReels-V2 and MAGI-1 demonstrate that MotionCache achieves significant speedups.
  arXiv  Detail & Related papers  (2026-05-03T05:49:27Z)
• WorldCache: Content-Aware Caching for Accelerated Video World Models [50.7543797435026]
  We present textbfWorldCache, a Perception-Constrained Dynamic CaCache framework.<n>WorldCache improves both when and how to reuse features.<n>On Cosmos-2.5-2B evaluated on PAI-Bench, WorldCache achieves textbf$2.3times$ inference speedup while preserving textbf99.4% of baseline quality.
  arXiv  Detail & Related papers  (2026-03-23T17:59:54Z)
• PreciseCache: Precise Feature Caching for Efficient and High-fidelity Video Generation [35.47114707080758]
  High computational costs and slow inference hinder the practical application of video generation models.<n>We propose textbfPreciseCache, a plug-and-play framework that precisely detects and skips truly redundant computations.
  arXiv  Detail & Related papers  (2026-03-01T08:08:49Z)
• SenCache: Accelerating Diffusion Model Inference via Sensitivity-Aware Caching [75.02865981328509]
  Caching reduces computation by reusing previously computed model outputs across timesteps.<n>We propose Sensitivity-Aware Caching (SenCache), a dynamic caching policy that adaptively selects caching timesteps on a per-sample basis.<n>SenCache achieves better visual quality than existing caching methods under similar computational budgets.
  arXiv  Detail & Related papers  (2026-02-27T17:36:09Z)
• Denoising as Path Planning: Training-Free Acceleration of Diffusion Models with DPCache [8.614492355393578]
  We propose DPCache, a training-free acceleration framework that formulates diffusion acceleration as a global path planning problem.<n> DPCache employs dynamic programming to select an optimal sequence of key timesteps that minimizes the total path cost while preserving trajectory fidelity.<n>Experiments on DiT, FLUX, and HunyuanVideo demonstrate that DPCache achieves strong acceleration with minimal quality loss.
  arXiv  Detail & Related papers  (2026-02-26T06:13:33Z)
• Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model [55.64316746098431]
  Timestep Embedding Aware Cache (TeaCache) is a training-free caching approach that estimates and leverages the fluctuating differences among model outputs across timesteps.<n>TeaCache achieves up to 4.41x acceleration over Open-Sora-Plan with negligible degradation of visual quality.
  arXiv  Detail & Related papers  (2024-11-28T12:50:05Z)
• DeepCache: Accelerating Diffusion Models for Free [65.02607075556742]
  DeepCache is a training-free paradigm that accelerates diffusion models from the perspective of model architecture. DeepCache capitalizes on the inherent temporal redundancy observed in the sequential denoising steps of diffusion models. Under the same throughput, DeepCache effectively achieves comparable or even marginally improved results with DDIM or PLMS.
  arXiv  Detail & Related papers  (2023-12-01T17:01:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.