UniVLR: Unifying Text and Vision in Visual Latent Reasoning for Multimodal LLMs

TL;DR

UniVLR introduces a unified visual reasoning framework that consolidates textual and visual reasoning into a shared visual workspace, improving efficiency and performance in multimodal large language models.

Contribution

It proposes a novel unified visual latent reasoning approach that replaces fragmented reasoning paths with a shared visual workspace, enhancing efficiency and reasoning quality.

Findings

Outperforms prior visual latent reasoning methods on perception and visual reasoning tasks.

Uses fewer reasoning tokens while maintaining or improving accuracy.

Reasoning is conducted solely through visual latents, eliminating external tool calls.

Abstract

Multimodal large language models are increasingly expected to perform thinking with images, yet existing visual latent reasoning methods still rely on explicit textual chain-of-thought interleaved with visual latent tokens. This interleaved design limits efficiency and keeps reasoning fragmented across separate text and vision channels. We propose UniVLR, a unified visual latent reasoning framework that treats textual reasoning and auxiliary visual evidence as a shared visual workspace. Instead of preserving text CoT as an independent inference-time path, UniVLR renders reasoning traces together with auxiliary images and learns to compress this unified representation into compact visual latent tokens. At inference time, the model reasons only through visual latents and directly decodes the final answer, avoiding both external tool calls and verbose text reasoning. Experiments on real-world perception and visual reasoning tasks show that UniVLR outperforms prior visual latent reasoning methods while using substantially fewer generated reasoning tokens, suggesting a more unified and efficient paradigm for visual thinking in MLLMs.