跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.24793 · v1

The Elephant in the Question: Unified Models Picture the Concepts Behind Their Answers

问题中的大象:统一模型描绘其答案背后的概念

首次提交 2026-11-19 01:38
版本更新 2026-11-19 01:38
雷达观测 2026-11-19 15:04
arXiv ↗PDF ↗

摘要

Visualizing intermediate representations could help us inspect the reasoning steps underlying a model output. Unified multimodal models offer this opportunity by combining understanding and generation in one architecture. Their image generators can access representations formed while processing text. We introduce the V-lens, which uses this connection to visualize a chosen token's hidden state without training a decoder. Across four diffusion-based and two purely autoregressive unified models, the resulting images reveal key intermediate concepts that connect a question to its answer even when the concepts are never named explicitly. Controlled tests show how this content changes with context, token position, and depth. We then introduce lens tuning, which supervises selected hidden states with pictures through the V-lens. Across BAGEL, SenseNova-U1, and HunyuanImage-3.0, the rate of correct visualizations of pronoun referents rises from 2-11% to 86-94%. Surprisingly, such visual supervision also improves text answers without supervising the answer tokens directly. Code is available at https://github.com/multimodal-ai-lab/V-lens.

作者备注:32 pages, including supplementary material. Code: https://github.com/multimodal-ai-lab/V-lens

AI 论文解读

基于摘要 · 非全文审阅

提出无需训练解码器的 V-lens,利用统一多模态模型的图像生成器可视化指定 token 的隐藏状态,并通过 lens tuning 显著提升代词指代可视化正确率。

正在读取账号与研究记录…

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-19 01:38 · 最新观测版本

The Elephant in the Question: Unified Models Picture the Concepts Behind Their Answers

Visualizing intermediate representations could help us inspect the reasoning steps underlying a model output. Unified multimodal models offer this opportunity by combining understanding and generation in one architecture. Their image generators can access representations formed while processing text. We introduce the V-lens, which uses this connection to visualize a chosen token's hidden state without training a decoder. Across four diffusion-based and two purely autoregressive unified models, the resulting images reveal key intermediate concepts that connect a question to its answer even when the concepts are never named explicitly. Controlled tests show how this content changes with context, token position, and depth. We then introduce lens tuning, which supervises selected hidden states with pictures through the V-lens. Across BAGEL, SenseNova-U1, and HunyuanImage-3.0, the rate of correct visualizations of pronoun referents rises from 2-11% to 86-94%. Surprisingly, such visual supervision also improves text answers without supervising the answer tokens directly. Code is available at https://github.com/multimodal-ai-lab/V-lens.

查看此版本解读 →查看此版本原文 ↗