---
title: "Visual Enhanced Depth Scaling for Multimodal Latent Reasoning"
created: 2026-04-15
updated: 2026-04-15
type: concept
tags: [multimodal, latent-reasoning, depth-scaling, vision]
sources: [raw/papers/2026/04/2604.10500.md]
---
# Visual Enhanced Depth Scaling

## 核心
[[multimodal-latent-reasoning]] 的后续工作：在潜在推理中引入视觉增强的深度缩放。

## 关键改进
- 增加视觉条件的推理深度
- 视觉 token 在 latent space 中获得更多推理层

## Related
- [[multimodal-latent-reasoning]]
- [[thinking-budget]]
- [[perception-grounded-po]]
