# Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

arXiv: https://arxiv.org/abs/2609.09143 · HTML: https://mlsanity.com/paper/2609.09143/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 63 of its 72 references that are in the map.
- First posted on arXiv 8 Sep 2026.

## Builds on (most cited first, 30 of 63)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [Image-to-Image Translation with Conditional Adversarial Networks](https://mlsanity.com/paper/1611.07004/) (2016)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [Taming Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2012.09841/) (2020)
- [Training Compute-Optimal Large Language Models](https://mlsanity.com/paper/2203.15556/) (2022)
- [Improved Techniques for Training GANs](https://mlsanity.com/paper/1606.03498/) (2016)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction](https://mlsanity.com/paper/2404.02905/) (2024)
- [Chameleon: Mixed-Modal Early-Fusion Foundation Models](https://mlsanity.com/paper/2405.09818/) (2024)
- [A Survey on Multimodal Large Language Models](https://mlsanity.com/paper/2306.13549/) (2023)
- [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025)
- [Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling](https://mlsanity.com/paper/2501.17811/) (2025)
- [Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation](https://mlsanity.com/paper/2406.06525/) (2024)
- [Vector-quantized Image Modeling with Improved VQGAN](https://mlsanity.com/paper/2110.04627/) (2021)
- [DataComp: In search of the next generation of multimodal datasets](https://mlsanity.com/paper/2304.14108/) (2023)
- [Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model](https://mlsanity.com/paper/2408.11039/) (2024)
- [LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset](https://mlsanity.com/paper/2309.11998/) (2023)
- [DataComp-LM: In search of the next generation of training sets for language models](https://mlsanity.com/paper/2406.11794/) (2024)
- [Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models](https://mlsanity.com/paper/2501.01423/) (2025)
- [BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset](https://mlsanity.com/paper/2505.09568/) (2025)

## Abstract (arXiv)

Image tokenizers define the ``visual language'' of unified multimodal models, yet are commonly studied through isolated metrics or generation-/understanding-only evaluations. These evaluations do not fully capture how visual tokens behave when modeled jointly with text. We build a controlled pure-autoregressive testbed and track task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. We examine how these losses scale and relate to downstream performance, then use them to study multimodal learnability---how well image and text tokens are jointly modeled---and tokenizer design. We find that (1) losses should be analyzed by task, since they exhibit distinct scaling behavior and rank tokenizers differently. (2) The loss--performance relationship depends on the predicted token space: for a fixed tokenizer, T2I and I2T losses correlate with generation quality, but across tokenizers, the T2I loss--performance relationship shifts with the image-token space, whereas I2T loss, computed over a shared text vocabulary, provides a more consistent signal. I2T loss also correlates with both generation and visual understanding performance after supervised finetuning. Using losses as a lens, we show that (3) better reconstruction does not necessarily yield lower task-specific losses or stronger downstream performance, and that (4) image tokenizer choice can affect text modeling under joint optimization. As case studies, we revisit three tokenizer design axes---the discriminator, semantic supervision, and vocabulary size---to examine their effects on joint modeling and downstream performance. Together, our testbed offers a complementary perspective on image tokenizers as visual languages, highlighting their interplay with text in joint multimodal training.