AI 连验证码都过不了
人类几乎闭眼能过的空间验证码,最强的 AI 还错四分之一。研究者造了 125 个要「点、拖、看结果再改」的网页验证码,考的是从 2D 画面推断 3D 布局、预判动作后果、失手后自我纠正——正是人拖动屏幕时脑子在做的事。人解出 94.1%,GPT-6 Astra 错 25.9%,最好的开源模型 Kimi K3 错 68%。更值得记的是失败方式:多数错在「看错截图」,而且模型一旦判断错了,推理很少回头改,反而编出证据替自己圆。他们顺手把这套题当训练场微调了一个小模型,成功率从 1.6% 拉到 33.1%,连真实室内视频的测试也涨了近 10 分。它不是你明天能用上的东西,但它说明一件事:会聊天不等于会看、会动手。
📄 原文摘要(英文)
Humans act in the visual world by inferring the 3D layout behind a 2D view, predicting how an action changes it, and correcting themselves after seeing the result. Can multimodal large language models (MLLMs) do the same? CAPTCHAs offer a natural test, since humans solve them through this very loop, clicking and dragging on the screen and adjusting after each result. Yet existing CAPTCHA benchmarks are largely solved and rarely spatial, while spatial benchmarks rarely require action. We therefore present TuringSpace, a synthetic web environment of 125 interactive CAPTCHAs that span five spatial abilities and must be solved through clicks and drags on the screen. Every task is a program that generates new instances with a checker and an oracle, so the tasks also form a scalable training gym. Humans solve 94.1% of the tasks, whereas the best model, GPT-6 Astra, still fails 25.9% of them, and the best open-weight model, Kimi K3, fails 68.0%. Our further analysis shows that most failures come from misreading the screenshot, and that model reasoning rarely corrects a wrong belief but hallucinates evidence to defend it. To test whether such spatial abilities can be learned, we fine-tune Qwen3.5-9B on this gym. Its success rate rises from 1.6% to 33.1%, above Kimi K3, and it improves on all seven out-of-distribution benchmarks, including a 9.9-point gain on the real-world indoor videos of VSI-Bench. These results show that grounding spatial intelligence in screen actions exposes what MLLMs lack and offers a scalable way to teach it, with gains that transfer to real-world scenes. We will release TuringSpace to support further study.