UniT: Unified Multimodal Chain-of-Thought Test-time Scaling

TL;DR

UniT introduces a unified multimodal reasoning framework that leverages test-time scaling and iterative reasoning to improve complex visual understanding and generation tasks, demonstrating better scalability and out-of-distribution performance.

Contribution

The paper presents UniT, a novel framework enabling iterative multimodal reasoning and verification within a single model, extending test-time scaling to multimodal tasks.

Findings

Unified models trained on short reasoning trajectories generalize to longer chains.

Sequential chain-of-thought reasoning is more scalable and compute-efficient than parallel sampling.

Training on generation and editing trajectories enhances out-of-distribution visual reasoning.

Abstract

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.