---
format: "aidr-story-markdown/v1"
id: "3fe55aec5351647bed9c4a2b00322d79f827d9d27bec9c8c5ba6a5ebe8c39efc"
canonical_url: "https://aidr.today/3fe55aec?lang=en"
title: "dNaViT: Discrete Codes are General-Purpose Visual Representations"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-11T00:00:00.000Z"
category: "Research"
topics: ["llm"]
source_urls: ["https://huggingface.co/papers/2611.12426"]
summary: "dNaViT is a Discrete Native-resolution Vision Transformer that produces discrete visual primitives analogous to language tokens. This work extends discrete autoregressive modeling to vision to foster a unified framework for visual and language models. dNaViT functions as a versatile discrete visual tokenizer enabling unified encoding across downstream tasks."
---

# dNaViT: Discrete Codes are General\-Purpose Visual Representations

> [Open the canonical story](<https://aidr.today/3fe55aec?lang=en>)

**Published:** 2026-11-11T00:00:00.000Z
**Category:** Research
**Topics:** llm

## Summary

dNaViT is a Discrete Native\-resolution Vision Transformer that produces discrete visual primitives analogous to language tokens\. This work extends discrete autoregressive modeling to vision to foster a unified framework for visual and language models\. dNaViT functions as a versatile discrete visual tokenizer enabling unified encoding across downstream tasks\.

## Sources

- [Story source](<https://huggingface.co/papers/2611.12426>)