AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Concept-Guided Tokenization: Closing the Gap Between Reconstruction and Generation

Yunqiao Yang, Haokun Lin, Guanzhong Wu, Ying Wei

ICML 2026 · Image generation › Autoregressive and latent generation · approach: Generative latent space tooling

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
add concept guidance (SAE concept-index loss) to the tokenizer (ablation)ConceptTok-B-B-64 baseline · ImageNet C2I (LlamaGen-B)gFID: 7.31 -> 6.51Improved
add text conditioning in the tokenizer encoder (ablation)ConceptTok-B-B-64 baseline · ImageNet C2I (LlamaGen-B)gFID: 7.31 -> 5.31Improved
combine text conditioning and concept guidance (ablation)ConceptTok-B-B-64 · ImageNet C2I (LlamaGen-B)gFID: 5.31 -> 4.13Improved
use concept guidance instead of holistic cosine alignment (ablation)ConceptTok-B-B-64 with text conditioning · ImageNet C2I (LlamaGen-B)gFID: 4.84 -> 4.13Improved
use concept guidance instead of cosine alignment (ablation)ConceptTok-B-B-64 · ImageNet CKNNACKNNA: 0.39 -> 0.48Improved
use concept-space alignment instead of dense feature alignment (REPA) for the diffusion generatorSiT-L/2 · ImageNet C2I 256x256, 400K iterationsgFID: 9.7 -> 8.7Improved
increase number of latent tokens from 32 to 128 (ablation)ConceptTok-B-L · ImageNet C2I (LlamaGen-B)gFID: 6.19 -> 2.97Improved
increase tokenizer size from S-S to B-B (64 tokens) (ablation)ConceptTok · ImageNet C2I (LlamaGen-B)gFID: 7.13 -> 4.13Improved
increase concept loss weight lambda from 0.1 to 0.5 (ablation)ConceptTok-B-B-64 · ImageNet C2I (LlamaGen-B)gFID: 4.13 -> 4.84Hurt
add concept guidance with lambda 0.1 versus lambda 0 (ablation)ConceptTok-B-B-64 · ImageNet val reconstructionrFID: 4.66 -> 3.84Improved
use CLIP/B instead of SigLIP/B as the pre-trained model for the SAE (ablation)ConceptTok-S-S-64 · ImageNet C2I (LlamaGen-B)gFID: 7.13 -> 6.80No clear effect
replace class-template captions with descriptive captions in training (ablation)ConceptTok-B-L-128 · COCO-30k T2IgFID: 10.73 -> 10.16Improved

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