Research map / Vision-language and multimodal / Multimodal LLMs
Liquid: Language Models are Scalable and Unified Multi-Modal Generators
Junfeng Wu, Yi Jiang, Chuofan Ma, Yuliang Liu, Hengshuang Zhao, Zehuan Yuan, Song Bai, Xiang Bai
Read the paperCheck your idea against 137,217 papers
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add 10M more visual generation data (ablation) | Liquid pretraining · VQAv2 | accuracy: 60.7 -> 64.5 | Improved |
| add 10M more visual generation data (ablation) | Liquid pretraining · MME | score: 909.6 -> 1066.8 | Improved |
| add 10M more visual understanding data (ablation) | Liquid pretraining · gFID | FID: 19.9 -> 14.9 | Improved |
| add 10M more generation data (ablation) | Liquid pretraining · gFID | FID: 19.9 -> 12.8 | Improved |
| add 10M more understanding data (ablation) | Liquid pretraining · GenAI-Bench overall | VQAScore: 0.60 -> 0.66 | Improved |
| train on mixed text and image data instead of text only (Gemma-7B) (ablation) | Liquid · text benchmarks MMLU | accuracy: 56.0 vs 55.5; HellaSwag 76.1 vs 77.2 | Mixed |
| train on mixed multimodal data instead of text-only (Llama3-1B) (ablation) | Liquid · HellaSwag | accuracy: 62.7 -> 55.0 | Hurt |
| train on mixed multimodal data instead of text-only (Qwen2.5-32B) (ablation) | Liquid · HellaSwag | accuracy: 79.2 -> 79.7 | No clear effect |
| train on mixed multimodal data instead of text-only (Qwen2.5-7B) (ablation) | Liquid · MMLU | accuracy: 64.9 -> 62.0 | Hurt |
| longer pretraining by one more epoch (ablation) | Liquid Gemma-7B · VQAv2 | accuracy: 68.0 -> 71.3 | Improved |
| replace Chameleon VQVAE with semantically aligned UniTok (ablation) | Liquid · MME | score: 1119.3 -> 1448.0 | Improved |
| replace Chameleon VQVAE with semantically aligned UniTok (ablation) | Liquid · TextVQA | accuracy: 42.4 -> 51.6 | Improved |
Later papers that compared with this method
1 later accepted paper reports Liquid (7B) as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Uni-X: Mitigating Modality Conflict with a Two-End-Separated Architecture for Unified Multimodal Models (ICLR 2026 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- VILA-U (256) — VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation (ICLR 2025) · better on 3 of 3 shared results
- Show-o — Show-o: One Single Transformer to Unify Multimodal Understanding and Generation (ICLR 2025) · better on 3 of 3 shared results
- Janus — Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation (CVPR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LMFusion: Adapting Pretrained Language Models for Multimodal Generation (NeurIPS 2025 · similarity 0.92)
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations (NeurIPS 2025 · similarity 0.91)
- Unified Language-Vision Pretraining in LLM with Dynamic Discrete Visual Tokenization (ICLR 2024 · similarity 0.91)
- MANZANO: A Simple and Scalable Unified Multimodal Model with a Hybrid Vision Tokenizer (ICLR 2026 · similarity 0.90)
- Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens (CVPR 2025 · similarity 0.90)
- Predictive Regularization Against Visual Representation Degradation in Multimodal Large Language Models (CVPR 2026 · similarity 0.90)
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.
