Image Reconstruction on ImageNet-1k 512x512 resolution (val)
0.4rFIDIBQ
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| IBQType=2D2026.04 | 0.4 | 25.19 | 0.662 | 0.201 | |
| MergeTok-BL (VAE)Type=1D VAE+VQ, Ratio=16, #Tok.=256, #Code=2^142026.05 | 0.42 | 21.8 | 0.69 | — | |
| MAGVIT-v2Type=2D2026.04 | 0.5 | 24.67 | 0.651 | 0.209 | |
| VibeToken-LLType=1D2026.04 | 0.51 | 23.37 | 0.7649 | 0.1867 | |
| VibeToken-SLType=1D2026.04 | 0.55 | 22.86 | 0.7541 | 0.1998 | |
| VAEType=2D VAE, Ratio=8, #Tok.=32^2, #Code=-2026.05 | 0.62 | 24.3 | 0.71 | — | |
| LLaMAGenType=2D2026.04 | 0.65 | 23.76 | 0.614 | 0.218 | |
| LlamaGen#Tokens=1024, Codebook Size=163842026.01 | 0.7 | 23.03 | — | — | |
| SoftVQ-BLType=1D VQ, Ratio=32, #Tok.=64, #Code=-2026.05 | 0.71 | — | — | — | |
| STATInference Strategy=threshold 0.01; +20 tokens, #Tokens=501, Codebook Size=40962026.01 | 0.75 | 20.88 | — | — | |
| STATInference Strategy=threshold 0.01, #Tokens=482, Codebook Size=40962026.01 | 0.82 | 20.98 | — | — | |
| MergeTok-BL (VQ)Type=1D VAE+VQ, Ratio=16, #Tok.=256, #Code=2^142026.05 | 0.82 | 21 | 0.67 | — | |
| STATInference Strategy=threshold 0.5, #Tokens=471, Codebook Size=40962026.01 | 0.87 | 20.98 | — | — | |
| InstellaType=1D2026.04 | 1.32 | 22.25 | 0.704 | — | |
| TiTok-B-128#Tokens=128, Codebook Size=40962026.01 | 1.52 | — | — | — | |
| TiTok-BType=1D2026.04 | 1.52 | — | — | — | |
| TiTok-L-64#Tokens=64, Codebook Size=40962026.01 | 1.77 | — | — | — | |
| MaskGIT VQ-GAN#Tokens=1024, Codebook Size=10242026.01 | 1.97 | — | — | — |