Latent Implicit Visual Reasoning

TL;DR

This paper introduces a task-agnostic method for Large Multimodal Models to discover and utilize visual reasoning tokens without explicit supervision, enhancing their ability to handle diverse vision-centric tasks.

Contribution

It proposes a novel, supervision-free visual reasoning mechanism that improves multimodal reasoning and generalizes across multiple vision tasks.

Findings

Outperforms direct fine-tuning on vision tasks

Achieves state-of-the-art results on diverse vision-centric tasks

Generalizes well to multi-task instruction tuning

Abstract

While Large Multimodal Models (LMMs) have made significant progress, they remain largely text-centric, relying on language as their core reasoning modality. As a result, they are limited in their ability to handle reasoning tasks that are predominantly visual. Recent approaches have sought to address this by supervising intermediate visual steps with helper images, depth maps, or image crops. However, these strategies impose restrictive priors on what "useful" visual abstractions look like, add heavy annotation costs, and struggle to generalize across tasks. To address this critical limitation, we propose a task-agnostic mechanism that trains LMMs to discover and use visual reasoning tokens without explicit supervision. These tokens attend globally and re-encode the image in a task-adaptive way, enabling the model to extract relevant visual information without hand-crafted supervision. Our approach outperforms direct fine-tuning and achieves state-of-the-art results on a diverse range of vision-centric tasks -- including those where intermediate abstractions are hard to specify -- while also generalizing to multi-task instruction tuning.