UniT: Unified Multimodal Chain-of-Thought Test-time Scaling
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914326342795264 |
|---|---|
| author | Chen, Leon Liangyu Ma, Haoyu Fan, Zhipeng Huang, Ziqi Sinha, Animesh Dai, Xiaoliang Wang, Jialiang He, Zecheng Yang, Jianwei Li, Chunyuan Sun, Junzhe Wang, Chu Yeung-Levy, Serena Juefei-Xu, Felix |
| author_facet | Chen, Leon Liangyu Ma, Haoyu Fan, Zhipeng Huang, Ziqi Sinha, Animesh Dai, Xiaoliang Wang, Jialiang He, Zecheng Yang, Jianwei Li, Chunyuan Sun, Junzhe Wang, Chu Yeung-Levy, Serena Juefei-Xu, Felix |
| contents | Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs. Many multimodal tasks, especially those involving complex spatial compositions, multiple interacting objects, or evolving instructions, require decomposing instructions, verifying intermediate results, and making iterative corrections. While test-time scaling (TTS) has demonstrated that allocating additional inference compute for iterative reasoning substantially improves language model performance, extending this paradigm to unified multimodal models remains an open challenge. We introduce UniT, a framework for multimodal chain-of-thought test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds. UniT combines agentic data synthesis, unified model training, and flexible test-time inference to elicit cognitive behaviors including verification, subgoal decomposition, and content memory. Our key findings are: (1) unified models trained on short reasoning trajectories generalize to longer inference chains at test time; (2) sequential chain-of-thought reasoning provides a more scalable and compute-efficient TTS strategy than parallel sampling; (3) training on generation and editing trajectories improves out-of-distribution visual reasoning. These results establish multimodal test-time scaling as an effective paradigm for advancing both generation and understanding in unified models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_12279 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | UniT: Unified Multimodal Chain-of-Thought Test-time Scaling Chen, Leon Liangyu Ma, Haoyu Fan, Zhipeng Huang, Ziqi Sinha, Animesh Dai, Xiaoliang Wang, Jialiang He, Zecheng Yang, Jianwei Li, Chunyuan Sun, Junzhe Wang, Chu Yeung-Levy, Serena Juefei-Xu, Felix Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs. Many multimodal tasks, especially those involving complex spatial compositions, multiple interacting objects, or evolving instructions, require decomposing instructions, verifying intermediate results, and making iterative corrections. While test-time scaling (TTS) has demonstrated that allocating additional inference compute for iterative reasoning substantially improves language model performance, extending this paradigm to unified multimodal models remains an open challenge. We introduce UniT, a framework for multimodal chain-of-thought test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds. UniT combines agentic data synthesis, unified model training, and flexible test-time inference to elicit cognitive behaviors including verification, subgoal decomposition, and content memory. Our key findings are: (1) unified models trained on short reasoning trajectories generalize to longer inference chains at test time; (2) sequential chain-of-thought reasoning provides a more scalable and compute-efficient TTS strategy than parallel sampling; (3) training on generation and editing trajectories improves out-of-distribution visual reasoning. These results establish multimodal test-time scaling as an effective paradigm for advancing both generation and understanding in unified models. |
| title | UniT: Unified Multimodal Chain-of-Thought Test-time Scaling |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2602.12279 |