Google DeepMind 与 Stanford 研究：图片符号能隐秘操纵模型写递归代码
DeepMind 和 Stanford 发现一张红心图片就能让 GPT-5.4 百分百写递归代码，而且思维链里完全看不出来，做 agent 的都该看看。
Google DeepMind 和 Stanford 的论文发现，在上下文里放一张红色爱心图片，Claude Opus 4.7 和 GPT-5.4 在中序二叉树遍历题上 100% 选择递归写法。一个黄色对勾符号能让 Gemini 3.1 Pro 在 90% 的样本中倾向递归，效果接近直接口头指令。同一符号的文字描述引导作用弱得多，随机或自然图片也更弱。GPT-5.4 的思维链中 74.7% 给出的解释与实际行为不符，模型无法说明原因。
Super interesting paper from Google DeepMind and Stanford.
They find that if you put a picture of a red heart in the context, Claude Opus 4.7 and GPT-5.4 write recursive code 100% of the time.
They found this on inorder binary tree traversal, a problem with equally valid recursive and iterative solutions. A yellow checkmark pushes Gemini 3.1 Pro to recursion in 90% of samples.
The effect matches or comes close to telling the model directly to use recursion. Text captions of the same symbol steer much less, and random or naturalistic images steer less still.
The models cannot explain it. Their interpretations of the symbol and their thinking traces do not match what they do. For GPT-5.4, 74.7% of traces give the wrong reason.
When models send images to each other, they pick shapes that they and humans find intuitive, and those shapes often fail to steer the other model.
If your agents read screenshots or images, those inputs can change behavior in ways the reasoning trace will not show.
https://t.co/qSlNRZmago
Chat with Paper: https://t.co/VGKo6Vhfg7
