TuringSpace: Grounding Spatial Intelligence in Action with Interactive CAPTCHAs
TuringSpace: Grounding Spatial Intelligence in Action with Interactive CAPTCHAs
Humans act in the visual world by inferring the 3D layout behind 2D views, predicting changes in actions, and correcting results. TuringSpace is a synthetic network environment containing 125 interactive CAPTCHAs, covering five spatial abilities, requiring screen clicks and drags to solve, with a human success rate of 94.1%, while GPT-6 Astra (the best model) still has a 25.9% task failure rate, and Kimi K3 (the best open-source model) has a failure rate as high as 68.0%.
Published on Nov 19
Authors:
,
,
,
,
,
,
,
,
,
,
Abstract
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.
View arXiv page View PDF Project page GitHub 0 Add to collection
Get this paper in your agent:
hf papers read 2611.25387
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash
Models citing this paper 0
No model linking this paper
Cite arxiv.org/abs/2611.25387 in a model README.md to link it from this page.
Datasets citing this paper 0
No dataset linking this paper
Cite arxiv.org/abs/2611.25387 in a dataset README.md to link it from this page.
Spaces citing this paper 0
No Space linking this paper
Cite arxiv.org/abs/2611.25387 in a Space README.md to link it from this page.
Collections including this paper 0
No Collection including this paper
Add this paper to a collection to link it from this page.
Source: Hugging Face Daily Papers · huggingface.co