DTC: Dependency-Aware Cache Execution and Deferred Trajectory Correction

Abstract Learned feature caching reduces diffusion model inference latency by replacing full denoising forwards with low-cost predictions. We propose a joint approach combining Dependency-Aware Cache Execution (DACE) and Deferred Trajectory Correction (DTC). When cache decomposition, prediction, and output projection are token-independent, DACE computes features only for target tokens and manages history liveness by future read positions, reducing cache-path computation and storage. DTC reuses scheduled full steps to measure endpoint counterfactual defects before old history refresh, approximates the error integral using signed sampler step sizes and cache age, and applies correction after normal state updates. A unified dependency plan makes execution pruning compatible with feedback-required history; both layers preserve model weights and the scheduled full DiT call budget. On Qwen-Image-Edit background-replacement tasks, the full method reduces GPU-resident N7 latency by 12.1% and mean LPIPS distance by 18.9% relative to LinCa; DTC adds 2.1% overhead to DACE while further reducing LPIPS by 18.6%.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11725324/v1
Primary Topic
Generative Adversarial Networks and Image Synthesis
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

DTC: Dependency-Aware Cache Execution and Deferred Trajectory Correction

Guandong Li, Mengxia Ye
1 links
Research Square
Generative Adversarial Networks and Image Synthesis
preprint

DTC: Dependency-Aware Cache Execution and Deferred Trajectory Correction

Guandong Li, Mengxia Ye
preprint en

Abstract

Abstract Learned feature caching reduces diffusion model inference latency by replacing full denoising forwards with low-cost predictions. We propose a joint approach combining Dependency-Aware Cache Execution (DACE) and Deferred Trajectory Correction (DTC). When cache decomposition, prediction, and output projection are token-independent, DACE computes features only for target tokens and manages history liveness by future read positions, reducing cache-path computation and storage. DTC reuses scheduled full steps to measure endpoint counterfactual defects before old history refresh, approximates the error integral using signed sampler step sizes and cache age, and applies correction after normal state updates. A unified dependency plan makes execution pruning compatible with feedback-required history; both layers preserve model weights and the scheduled full DiT call budget. On Qwen-Image-Edit background-replacement tasks, the full method reduces GPU-resident N7 latency by 12.1% and mean LPIPS distance by 18.9% relative to LinCa; DTC adds 2.1% overhead to DACE while further reducing LPIPS by 18.6%.

Research Square
IFlyTek (China) (CN)
Generative Adversarial Networks and Image Synthesis
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon