{"paper":{"title":"Mogao: An Omni Foundation Model for Interleaved Multi-Modal Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mogao is a single model that generates arbitrary sequences mixing text and images by fusing autoregressive and diffusion components.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chao Liao, Jie Wu, Liang Li, Liyang Liu, Weilin Huang, Wenliang Zhao, Xinyu Zhang, Xun Wang, Zhengxiong Luo, Zhi Tian","submitted_at":"2025-05-08T17:58:57Z","abstract_excerpt":"Recent progress in unified models for image understanding and generation has been impressive, yet most approaches remain limited to single-modal generation conditioned on multiple modalities. In this paper, we present Mogao, a unified framework that advances this paradigm by enabling interleaved multi-modal generation through a causal approach. Mogao integrates a set of key technical improvements in architecture design, including a deep-fusion design, dual vision encoders, interleaved rotary position embeddings, and multi-modal classifier-free guidance, which allow it to harness the strengths "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Mogao not only achieves state-of-the-art performance in multi-modal understanding and text-to-image generation, but also excels in producing high-quality, coherent interleaved outputs. Its emergent capabilities in zero-shot image editing and compositional generation highlight Mogao as a practical omni-modal foundation model.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The key technical improvements (deep-fusion design, dual vision encoders, interleaved rotary position embeddings, and multi-modal classifier-free guidance) successfully combine the strengths of autoregressive text models and diffusion image models for arbitrary interleaved sequences.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Mogao presents a causal unified model with deep fusion, dual encoders, and interleaved position embeddings that achieves strong performance on multi-modal understanding, text-to-image generation, and coherent interleaved outputs including zero-shot editing.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Mogao is a single model that generates arbitrary sequences mixing text and images by fusing autoregressive and diffusion components.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bf1f45e253a85302aa1d5040e96a7df07fd3449de3d7b21a854d1d16a302836a"},"source":{"id":"2505.05472","kind":"arxiv","version":2},"verdict":{"id":"e4dc66b9-5e09-4357-a230-27a0a06abfe0","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T07:18:40.194337Z","strongest_claim":"Mogao not only achieves state-of-the-art performance in multi-modal understanding and text-to-image generation, but also excels in producing high-quality, coherent interleaved outputs. Its emergent capabilities in zero-shot image editing and compositional generation highlight Mogao as a practical omni-modal foundation model.","one_line_summary":"Mogao presents a causal unified model with deep fusion, dual encoders, and interleaved position embeddings that achieves strong performance on multi-modal understanding, text-to-image generation, and coherent interleaved outputs including zero-shot editing.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The key technical improvements (deep-fusion design, dual vision encoders, interleaved rotary position embeddings, and multi-modal classifier-free guidance) successfully combine the strengths of autoregressive text models and diffusion image models for arbitrary interleaved sequences.","pith_extraction_headline":"Mogao is a single model that generates arbitrary sequences mixing text and images by fusing autoregressive and diffusion components."},"references":{"count":99,"sample":[{"doi":"","year":2024,"title":"Pixtral 12b.arXiv preprint arXiv:2410.07073, 2024","work_id":"9ad2b071-82d8-4cfa-b994-b9975094b575","ref_index":1,"cited_arxiv_id":"2410.07073","is_internal_anchor":true},{"doi":"","year":2022,"title":"Flamingo: a visual language model for few-shot learning","work_id":"b61d581d-a5f8-4799-a638-d91ddfc06da4","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond","work_id":"cbc2bb21-b6bb-46c0-80bf-107e195ffe10","ref_index":3,"cited_arxiv_id":"2308.12966","is_internal_anchor":true},{"doi":"","year":2025,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","ref_index":4,"cited_arxiv_id":"2502.13923","is_internal_anchor":true},{"doi":"","year":2023,"title":"Improving image generation with better captions.Computer Science","work_id":"fb7509a6-ece6-4ea3-b583-1ec884016dc8","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":99,"snapshot_sha256":"08c6054537d19ec7e98dafbb8ea80acc44e066f2b817c034c113ca16062ba8d0","internal_anchors":37},"formal_canon":{"evidence_count":3,"snapshot_sha256":"b7903a9725fa117ff04df9c21506486f30477dfc034a2c0d69cced73bf9617af"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}