Search papers, labs, and topics across Lattice.
MentalThink introduces a visual-symbolic reasoning paradigm that enhances Multimodal LLMs (MLLMs) with an executable mechanism for mental visualization through a think-with-SVG pipeline. This method allows models to generate and interpret scalable vector graphics (SVG) as an intermediate representation, facilitating multi-turn reasoning and mimicking human mental imagery processes. Evaluations reveal that MentalThink significantly outperforms existing models on spatial understanding benchmarks, achieving scores of 55.1% on VSIBench and 76.0% on MindCube, highlighting the effectiveness of SVG as a visual workspace for reasoning tasks.
Executable vector graphics enable MLLMs to achieve human-like spatial reasoning through a structured visual workspace.
We introduce MentalThink, a visual-symbolic reasoning paradigm that equips Multimodal LLMs (MLLMs) with an executable mechanism for"mental"visualization. The core of MentalThink is a think-with-SVG pipeline, where the model learns to generate, render, and interpret scalable vector graphics (SVG) code as an intermediate visual representation for multi-turn reasoning. By creating structured vector sketches, the model can externalize spatial hypotheses, inspect them through deterministic rendering, and reason within a constrained geometric space, effectively mimicking the human process of mental imagery. We instantiate this paradigm through a two-stage training framework, combining Supervised Fine-Tuning (SFT) for SVG syntactic alignment with multi-turn Reinforcement Learning (RL) to encourage iterative inspection, revision, and refinement of intermediate visual hypotheses. Extensive evaluations demonstrate that MentalThink achieves superior performance on spatial understanding and reasoning benchmarks (e.g., 55.1% on VSIBench, 76.0% on MindCube), showing that executable vector graphics provide a verifiable visual workspace for dynamic perspective taking, visual reflection, and compositional scene construction.
