AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Scalable Adaptive Computation for Iterative Generation

Allan Jabri, David J. Fleet, Ting Chen

ICML 2023 · Image generation › Autoregressive and latent generation · approach: Generative latent space tooling · 14 citations (OpenAlex)

Read the paper

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
latent self-conditioning (training rate 0.9 vs 0) (ablation)RIN · ImageNet 64x64 and 128x128 ablationFID/IS: clear correlation between self-conditioning rate and sample qualityImproved
multiple read-write routing rounds (more stacked blocks) vs single read-write (ablation)RIN · ImageNet ablation over model sizeFID/IS: saturates later with significantly better sample qualityImproved
sigmoid noise schedule (temperature 0.9) instead of cosine for training (ablation)RIN · ImageNet larger resolutionstraining stabilityImproved
noise schedule choice at sampling time (sigmoid vs cosine) (ablation)RIN · ImageNet samplingFIDNo clear effect
large tokenization (1x1 patches, 4096 interface tokens) vs 8x8 patches (ablation)RIN · ImageNet ablationFID/ISMixed
input scalingRIN at 256x256 · ImageNet 256x256FID: 4.51 -> 3.42Improved
RIN architecture vs U-Net diffusion baselines with fewer FLOPsRIN · ImageNet 64x64 to 256x256FID and GFLOPs: e.g. 1.23 FID at 106 GFLOPs vs ADM 210 GFLOPsImproved
RIN vs CF-guidance at 128x128 without guidanceRIN · ImageNet 128x128FID: 2.75 vs 2.43Hurt
RIN without guidance vs ADM with guidance at 256x256RIN · ImageNet 256x256FID: RIN+inp.scale 3.42 vs ADM+guid 4.59 (better); RIN 4.51 vs ADM+guid 4.59 (roughly tied)Mixed
RIN video model vs Video Diffusion with reconstruction guidanceRIN · Kinetics-600 16x64x64FVD: 10.8 vs 16.6Improved
sampling steps 400 vs 1000 for RIN video (ablation)RIN · Kinetics-600 16x64x64FVD: 11.5 -> 10.8Improved

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