# Intuition vectors

arXiv: https://arxiv.org/abs/2609.32619 · HTML: https://mlsanity.com/paper/2609.32619/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 37 references that are in the map.
- First posted on arXiv 26 Sep 2026.

## Builds on (most cited first, 26 of 26)

- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets](https://mlsanity.com/paper/2310.06824/) (2023)
- [Designing and Interpreting Probes with Control Tasks](https://mlsanity.com/paper/1909.03368/) (2019)
- [DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning](https://mlsanity.com/paper/2411.04983/) (2024)
- [Diffusion Transformers with Representation Autoencoders](https://mlsanity.com/paper/2510.11690/) (2025)
- [Not All Language Model Features Are One-Dimensionally Linear](https://mlsanity.com/paper/2405.14860/) (2024)
- [ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems](https://mlsanity.com/paper/2505.11831/) (2025)
- [Hierarchical Reasoning Model](https://mlsanity.com/paper/2506.21734/) (2025)
- [The Geometry of Categorical and Hierarchical Concepts in Large Language Models](https://mlsanity.com/paper/2406.01506/) (2024)
- [Evaluating the World Model Implicit in a Generative Model](https://mlsanity.com/paper/2406.03689/) (2024)
- [The Surprising Effectiveness of Test-Time Training for Few-Shot Learning](https://mlsanity.com/paper/2411.07279/) (2024)
- [Bongard-LOGO: A New Benchmark for Human-Level Concept Learning and Reasoning](https://mlsanity.com/paper/2010.00763/) (2020)
- [Discovering Interpretable Directions in the Semantic Latent Space of Diffusion Models](https://mlsanity.com/paper/2303.11073/) (2023)
- [Revisiting the Platonic Representation Hypothesis: An Aristotelian View](https://mlsanity.com/paper/2602.14486/) (2026)
- [Into the Rabbit Hull: From Task-Relevant Concepts in DINO to Minkowski Geometry](https://mlsanity.com/paper/2510.08638/) (2025)
- [ARC Is a Vision Problem!](https://mlsanity.com/paper/2511.14761/) (2025)
- [Searching Latent Program Spaces](https://mlsanity.com/paper/2411.08706/) (2024)
- [Bongard-OpenWorld: Few-Shot Reasoning for Free-form Visual Concepts in the Real World](https://mlsanity.com/paper/2310.10207/) (2023)
- [Product of Experts with LLMs: Boosting Performance on ARC Is a Matter of Perspective](https://mlsanity.com/paper/2505.07859/) (2025)
- [Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objects](https://mlsanity.com/paper/2406.15955/) (2024)
- [Adapting Self-Supervised Representations as a Latent Space for Efficient Generation](https://mlsanity.com/paper/2510.14630/) (2025)
- [ARC-GEN: A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpus](https://mlsanity.com/paper/2511.00162/) (2025)
- [Support-Set Context Matters for Bongard Problems](https://mlsanity.com/paper/2309.03468/) (2023)
- [Unraveling the geometry of visual relational reasoning](https://mlsanity.com/paper/2502.17382/) (2025)

## Abstract (arXiv)

Large self-supervised vision models learn representations that support scene segmentation and the semantic decomposition of physical objects. We ask whether their representational geometry supports transfer to visual reasoning problems without any task-specific fine-tuning. We hypothesized that relational representations may bridge perception and abstract reasoning by encoding similarities and transformations among visual inputs such that an intuitive, implicit form of reasoning may be performed via latent vector arithmetic. Specifically, we examine DINOv3, MAE, and random pixel projections on abstract and naturalistic Bongard problems, ARC-AGI-1 and ARC-AGI-2, and novel ARC-GEN instances. On both Bongard benchmarks, the accuracy of a simple nearest-centroid readout of frozen visual embeddings is within four percentage points of the task-specific baselines reported with the original benchmarks. In ARC, latent difference vectors summarizing demonstration input-output transformations, which we refer to as intuition vectors, show greater alignment with test vectors from the same task, whereas those from unrelated tasks are near orthogonal. This latent geometry is operational: transporting a query along its intuition vector consistently improves exact-output retrieval, reaching 70.7 on ARC-AGI-2 evaluation. Across 397,000 ARC-GEN instances from 794 tasks, single-pair intuition vectors identify the generating task with approximately 87% leave-one-out accuracy. These findings suggest that latent vector arithmetic over frozen visual representations supports implicit rule inference across varied problem domains without a generative model component, indicating that inferring an abstract transformation and generating its instance-specific consequence may be separable capacities.