Google DeepMind 与 Stanford 论文：提示词中的红心图形可让 VLM 转向递归解法

Rohan Paul· @rohanpaul_ai · X·2026-11-04 00:14· 1 小时前AI 评分56

AI 导读

Google DeepMind 与 Stanford 的论文发现，提示词中加入与任务无关的简单彩色图形（如红心）会隐式改变 VLM 的行为。在二叉树编码题上，红心使 Claude Opus 4.7 和 GPT-5.4 的递归解法占比从 0% 升到 100%；Gemini 3.1 Pro 被图形影响时，71.5% 的思考痕迹称图像无关。

正文 · 原文

so interesting. 🤔

New Google DeepMind and Stanford paper finds that a red heart picture in the prompt can decide how AI write code, that simple shapes in images can steer AI on unrelated tasks, often without the models saying so.

plain colored shapes in a prompt change how AI models write code, though the shapes have nothing to do with it.

Vision-language models now read screenshots and images as part of normal work. If an unrelated picture changes the model's choice, builders need to see it, ideally in the model's own reasoning.

They added a plain colored shape, like a red heart, to a coding prompt and told the model to look at it. Across 288 shapes, they tracked how often Claude Opus 4.7, Gemini 3.1 Pro, and GPT-5.4 chose recursion.

A red heart moved Claude Opus 4.7 and GPT-5.4 from 0% recursive solutions to 100% on a binary tree problem. When a shape did steer Gemini 3.1 Pro, 71.5% of its thinking traces called the image irrelevant.

If your agent sees images, test its decisions with and without them, and do not count on reasoning traces to flag visual influence.

来源：Rohan Paul · x.com