AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding

Weichen Fan, Haiwen Diao, Quan Wang, Dahua Lin, Ziwei Liu

ECCV 2026 · Image generation › Autoregressive and latent generation · approach: Hierarchical VQ-VAE generation · 19 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
add semantic alignment loss on 25% of low-frequency tokens (ablation)UAE (DINOv2) · ImageNet-1K linear probingaccuracy: 60.1 -> 83.0Improved
add semantic alignment loss on 25% of low-frequency tokens (ablation)UAE (DINOv2) · ImageNet-1K reconstructionPSNR: 34.3 -> 32.2Hurt
increase semantic-loss fraction K_base from 25% to 75% (ablation)UAE (DINOv2) · ImageNet-1K reconstructionPSNR: 32.2 -> 20.1Hurt
mask 85% instead of 25% of high-frequency tokens (ablation)UAE · latent generationgFID@20 epoch: 33.1 -> 9.7Improved
mask 15% instead of 25% of high-frequency tokens (ablation)UAE · latent generationgFID@20 epoch: 33.1 -> 12.2Improved
keep only 64 low-frequency tokens instead of 121 (ablation)UAE latent diffusion · latent generation at 20 epochsFID: 40.5 -> 9.7Improved
keep only 64 low-frequency tokens when training from scratch (ablation)encoder-decoder from scratch · linear probingaccuracy: 20.7 -> 60.1Improved
apply unified training to RAE-style DINOv2-B tokenizerDINOv2-B tokenizer · ImageNet-1K reconstructionPSNR: 18.05 -> 31.00Improved
apply unified training to RAE-style DINOv2-B tokenizerDINOv2-B tokenizer · MS-COCO 2017 reconstructionrFID: 6.01 -> 2.01Improved
apply UAE unified training to JITJIT 700M, 80 epochs · pixel-space ImageNet generationFID: 16.06 -> 14.55Improved
apply UAE unified training to JITJIT 700M, 40 epochs · pixel-space ImageNet generationIS: 60.13 -> 72.82Improved
generate with 64 UAE low-frequency tokens instead of RAE latents in SiTSiT · inference costGFLOPs: 313.84 -> 78.80Improved

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