Latent Sketchpad: Sketching Visual Thoughts to Elicit Multimodal Reasoning in MLLMs

TL;DR

Latent Sketchpad introduces an internal visual scratchpad for MLLMs, enabling visual thinking that enhances reasoning and interpretability without sacrificing performance, demonstrated on the MazePlanning dataset.

Contribution

The paper presents a novel framework that integrates visual generation into MLLMs' reasoning process, supporting multimodal thought and improving interpretability.

Findings

Achieves comparable or superior reasoning performance to backbone models.

Generalizes across different MLLMs like Gemma3 and Qwen2.5-VL.

Enhances interpretability through visual latents and sketches.

Abstract

While Multimodal Large Language Models (MLLMs) excel at visual understanding, they often struggle in complex scenarios that require visual planning and imagination. Inspired by how humans use sketching as a form of visual thinking to develop and communicate ideas, we introduce Latent Sketchpad, a framework that equips MLLMs with an internal visual scratchpad. The internal visual representations of MLLMs have traditionally been confined to perceptual understanding. We repurpose them to support generative visual thought without compromising reasoning ability. Building on frontier MLLMs, our approach integrates visual generation directly into their native autoregressive reasoning process. It allows the model to interleave textual reasoning with the generation of visual latents. These latents guide the internal thought process and can be translated into sketch images for interpretability. To realize this, we introduce two components: a Context-Aware Vision Head autoregressively produces visual representations, and a pretrained Sketch Decoder renders these into human-interpretable images. We evaluate the framework on our new dataset MazePlanning. Experiments across various MLLMs show that Latent Sketchpad delivers comparable or even superior reasoning performance to their backbone. It further generalizes across distinct frontier MLLMs, including Gemma3 and Qwen2.5-VL. By extending model's textual reasoning to visual thinking, our framework opens new opportunities for richer human-computer interaction and broader applications. More details and resources are available on our project page: https://latent-sketchpad.github.io/.