AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Accelerating Controllable Generation via Hybrid-grained Cache

Lin Liu, Huixia Ben, Shuo Wang, Jinda Lu, Junxiang Qiu, Shengeng Tang, Yanbin Hao

AAAI 2026 · Image generation › Autoregressive and latent generation · approach: Compositional scene generation · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply HGC (block-level and prompt-level cache) instead of no cacheControlNet · ADE20k segmentationMACs: 18.22T -> 6.70TImproved
apply HGC instead of no cacheControlNet · ADE20k segmentationFID: 38.13 -> 39.44Hurt
apply HGC to ControlNet++ instead of no cacheControlNet++ · ADE20k segmentationFID: 30.51 -> 28.93Improved
apply HGC instead of DeepCache at similar MACsControlNet · COCO-Stuff segmentationFID: 24.84 -> 20.29Improved
lower similarity threshold theta from 0.9 to 0 (ablation)HGC block-level cache · ADE20k segmentationFID: 28.93 -> 30.92Hurt
raise threshold theta from 0.9 to 1.0 (ablation)HGC block-level cache · ADE20k segmentationFID: 28.93 -> 28.72No clear effect
increase lambda_intra from 0 to 1.0 (ablation)HGC generative module · ADE20k segmentationFID: 28.86 -> 29.04Hurt
increase lambda_inter from 0.6 to 1.0 (ablation)HGC generative module · ADE20k segmentationFID: 28.93 -> 30.49Hurt
increase lambda_inter from 0 to 0.6 (ablation)HGC generative module · ADE20k segmentationFID: 28.44 -> 28.93No clear effect
cache generative module blocks only (ablation)ControlNet++ · COCO-Stuff segmentationMACs: 18.22T -> 12.09TImproved
cache control module blocks only (ablation)ControlNet++ · COCO-Stuff segmentationMACs: 18.22T -> 14.95TImproved
apply HGC to video generation with CTRL-AdapterCTRL-Adapter · DAVIS 2017MACs and FID: 48.29T -> 28.88T MACs; FID 11.03 -> 11.70Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy