← back to paper
arxiv: 2606.27313 · 2 revisions
ViQ: Text-Aligned Visual Quantized Representations at Any Resolution