FeatFix: Reuse What You Verify through Local Exact-Feature Correction for Faster Cached Diffusion Inference

2026-07-30

AI Summary

This work addresses the high computational cost of iterative denoising in diffusion models, where existing acceleration methods discard precise features after verification and fail to effectively reuse prior computations. The authors propose a local exact feature correction mechanism that, at sparsely selected layer–timestep positions, directly replaces draft outputs with exact block outputs computed during the verification phase, thereby resetting local residuals to suppress error propagation. Requiring no additional training, this approach combines cache-based inference with a localized replacement strategy to achieve, for the first time, direct reuse of exact features—avoiding both partial substitutions and full-step recomputation. Evaluated across four image and video diffusion backbones, the method achieves up to 6.70× speedup while maintaining competitive generation quality.

Abstract

Diffusion models are widely used to generate high-quality images and videos, but their iterative denoising process remains computationally intensive. A growing class of training-free accelerators reduces this cost by reusing cached intermediate features or forecasting future ones. To control draft drift, these methods sometimes compute an exact block feature for verification. Yet the resulting exact feature is typically used only to measure discrepancy or guide a later decision and is then discarded. We find that this previously computed feature can instead be reused for correction. Forwarding it at the verification site resets the local draft residual and reduces downstream feature error. Based on this observation, we introduce FeatFix, a local exact-feature correction method for cached diffusion inference. FeatFix operates at a fixed sparse set of layer--timestep sites. At each selected site, it replaces the complete draft block output with the exact output computed from the same incoming state, avoiding token- or channel-level partial replacement and full-timestep recomputation. Experiments across four image and video backbones show that FeatFix consistently accelerates generation, achieving a speedup of up to $6.70\times$ over Vanilla while maintaining competitive output quality.

Problem

• diffusion models
• cached inference
• feature reuse
• acceleration
• exact-feature verification

Innovation

• cached diffusion inference
• exact-feature correction
• draft drift mitigation
• training-free acceleration
• local residual reset