AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Learning from Semantic Dictionaries: Discriminative Codebook Contrastive Learning for Unified Visual Representation and Generation

Imanol G. Estepa, Jesús M. Rodríguez-de-Vera, Bhalaji Nagarajan, Petia Radeva

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Hierarchical VQ-VAE generation · 0 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
add encoder codebook contrast to reconstruction objective (ablation)LEASE · IN-1K ablationlinear probe accuracy: 73.62 -> 76.11Improved
add encoder codebook contrast to reconstruction objective (ablation)LEASE · IN-1K ablationIS: 79.59 -> 86.71Improved
apply contrast on decoder features instead of encoder (ablation)LEASE · IN-1K ablationFID: 10.62 -> 10.97Mixed
add decoder contrast on top of encoder contrast (ablation)LEASE · IN-1K ablationlinear probe accuracy: 76.11 -> 76.07No clear effect
double codebook size from 8K to 16K centroids (ablation)LEASE · IN-1Klinear probe accuracy: 75.20 -> 76.70Improved
use DINOv2 codebook instead of DINOv3 on IN200 (ablation)LEASE · IN200 (200 ep)linear probe accuracy: 79.98 -> 85.86Improved
build DINOv2 codebook on IN200 instead of IN-1K (ablation)LEASE · IN200 (200 ep)linear probe accuracy: 85.47 -> 85.86Improved
raise contrast weight lambda from 0.1 to 1.0 (ablation)LEASE · IN200 (200 ep)linear probe accuracy: 85.47 -> 84.93Hurt
use 15 neighbor centroids instead of 5 (ablation)LEASE · IN200 (200 ep)FID: 16.33 -> 16.80Hurt
use 30 neighbor centroids instead of 5 (ablation)LEASE · IN200 (200 ep)linear probe accuracy: 85.47 -> 85.36No clear effect
LEASE versus MAGE in low-data linear probingMAGE · IN-1K few-shottop-1 accuracy: 57.33 -> 59.65Improved
LEASE versus MAGE under ImageNet-A shiftMAGE · ImageNet-A k-NNk-NN top-1 accuracy: 4.19 -> 8.51Improved

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