reasoning from minimal examples remains a core unsolved problem for frontier foundation models such as GPT-5 and Grok 4. These models still fail to infer structured transformation rules from a handful of examples, which is a key hallmark of human intelligence. The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) provides a rigorous testbed for this capability, demanding conceptual rule induction and transfer to novel tasks. Most existing methods treat ARC-AGI as a purely textual reasoning task, overlooking the fact that humans rely heavily on visual abstraction when solving such puzzles. However, our pilot experiments reveal a paradox: naively rendering ARC-AGI grids as images degrades performance due to imprecise rule execution. This leads to our central hypothesis that vision and language possess complementary strengths across distinct reasoning stages: vision supports global pattern abstraction and verification, whereas language specializes in symbolic rule formulation and precise execution. Building on this insight, we introduce two synergistic strategies: (1) Vision-Language Synergy Reasoning (VLSR), which decomposes ARC-AGI into modality-aligned subtasks; and (2) Modality-Switch Self-Correction (MSSC), which leverages vision to verify text-based reasoning for intrinsic error correction. Extensive experiments demonstrate that our approach yields up to a 4.33% improvement over text-only baselines across diverse flagship models and multiple ARC-AGI tasks. Our findings suggest that unifying visual abstraction with linguistic reasoning is a crucial step toward achieving general-izable, human-like intelligence in future foundation models. Source code is released at https://github.com/

The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) is a generative, few-shot fluid intelligence benchmark. Although humans effortlessly solve ARC-AGI, it remains extremely difficult for even the most advanced artificial intelligence systems. Inspired by methods for modelling human intelligence spanning neuroscience to psychology, we propose a cognitively plausible ARC-AGI solver. Our solver integrates System 1 intuitions with System 2 reasoning in an efficient and interpretable process using neurosymbolic methods based on Vector Symbolic Algebras (VSAs). Our solver works by object-centric program synthesis, leveraging VSAs to represent abstract objects, guide solution search, and enable sample-efficient neural learning. Preliminary results indicate success, with our solver scoring 10.8% on ARC-AGI-1-Train and 3.0% on ARC-AGI-1-Eval. Additionally, our solver performs well on simpler benchmarks, scoring 94.5% on Sort-of-ARC and 83.1% on 1D-ARC -- the latter outperforming GPT-4 at a tiny fraction of the computational cost. Importantly, our approach is unique; we believe we are the first to apply VSAs to ARC-AGI and have developed the most cognitively plausible ARC-AGI solver yet. Our code is available at: https://github.com/ijoffe/ARC-VSA-2025.

We run a controlled compositional generalization experiment in the ARC-AGI domain: an open-world problem domain in which the ability to generalize out-of-distribution is, by design, an essential characteristic for success. We compare neural program synthesis and test-time fine-tuning approaches on this experiment. We find that execution-guided neural program synthesis outperforms all reference algorithms in its ability to compose novel solutions. Our empirical findings also suggest that the success of TTFT on ARC-AGI lies mainly in eliciting in-distribution knowledge that the LLM otherwise fails to rely on directly.

The Abstraction and Reasoning Corpus (ARC-AGI) poses a significant challenge for large language models (LLMs), exposing limitations in their abstract reasoning abilities. In this work, we leverage task-specific data augmentations throughout the training, generation, and scoring phases, and employ a depth-first search algorithm to generate diverse, high-probability candidate solutions. Furthermore, we utilize the LLM not only as a generator but also as a scorer, using its output probabilities to select the most promising solutions. Our method achieves a score of 71.6% (286.5/400 solved tasks) on the public ARC-AGI evaluation set, demonstrating state-of-the-art performance among publicly available approaches. While concurrent closed-source work has reported higher scores, our method distinguishes itself through its transparency, reproducibility, and remarkably low inference cost, averaging only around 2ct per task on readily available hardware (we assume a price of 36ct/hour for a Nvidia 4090 GPU).

Video Diffusion Models (VDMs) have emerged as powerful generative tools, capable of synthesizing high-quality spatiotemporal content. Yet, their potential goes far beyond mere video generation. We argue that the training dynamics of VDMs, driven by the need to model coherent sequences, naturally pushes them to internalize structured representations and an implicit understanding of the visual world. To probe the extent of this internal knowledge, we introduce a few-shot fine-tuning framework that repurposes VDMs for new tasks using only a handful of examples. Our method transforms each task into a visual transition, enabling the training of LoRA weights on short input-output sequences without altering the generative interface of a frozen VDM. Despite minimal supervision, the model exhibits strong generalization across diverse tasks, from low-level vision (for example, segmentation and pose estimation) to high-level reasoning (for example, on ARC-AGI). These results reframe VDMs as more than generative engines. They are adaptable visual learners with the potential to serve as the backbone for future foundation models in vision.

They want to build AI models that achieve "artificial general intelligence," or AGI--matching or exceeding the capabilities of the human mind. The difference between these two men is that Altman has suggested that his company, OpenAI, has practically built the technology already. Chollet, a French computer scientist and one of the industry's sharpest skeptics, has said that notion is "absolutely clown shoes." When I spoke with him earlier this year, Chollet told me that AI companies have long been "intellectually lazy" in suggesting that their machines are on the path to a kind of supreme knowledge. At this point, those claims are based largely on the programs' ability to pass specific tests (such as the LSAT, Advanced Placement Biology, and even an introductory sommelier exam).

OpenAI's o3 achieves a high score of 87.5 % on ARC-AGI, a benchmark proposed to measure intelligence. This raises the question whether systems based on Large Language Models (LLMs), particularly o3, demonstrate intelligence and progress towards artificial general intelligence (AGI). Building on the distinction between skills and intelligence made by Fran\c{c}ois Chollet, the creator of ARC-AGI, a new understanding of intelligence is introduced: an agent is the more intelligent, the more efficiently it can achieve the more diverse goals in the more diverse worlds with the less knowledge. An analysis of the ARC-AGI benchmark shows that its tasks represent a very specific type of problem that can be solved by massive trialling of combinations of predefined operations. This method is also applied by o3, achieving its high score through the extensive use of computing power. However, for most problems in the physical world and in the human domain, solutions cannot be tested in advance and predefined operations are not available. Consequently, massive trialling of predefined operations, as o3 does, cannot be a basis for AGI - instead, new approaches are required that can reliably solve a wide variety of problems without existing skills. To support this development, a new benchmark for intelligence is outlined that covers a much higher diversity of unknown tasks to be solved, thus enabling a comprehensive assessment of intelligence and of progress towards AGI.

As of December 2024, the ARC-AGI benchmark is five years old and remains unbeaten. We believe it is currently the most important unsolved AI benchmark in the world because it seeks to measure generalization on novel tasks -- the essence of intelligence -- as opposed to skill at tasks that can be prepared for in advance. This year, we launched ARC Prize, a global competition to inspire new ideas and drive open progress towards AGI by reaching a target benchmark score of 85\%. As a result, the state-of-the-art score on the ARC-AGI private evaluation set increased from 33\% to 55.5\%, propelled by several frontier AGI reasoning techniques including deep learning-guided program synthesis and test-time training. In this paper, we survey top approaches, review new open-source implementations, discuss the limitations of the ARC-AGI-1 dataset, and share key insights gained from the competition.

ARC-AGI is an open-world problem domain in which the ability to generalize out-of-distribution is a crucial quality. Under the program induction paradigm, we present a series of experiments that reveal the efficiency and generalization characteristics of various neurally-guided program induction approaches. The three paradigms we consider are Learning the grid space, Learning the program space, and Learning the transform space. We implement and experiment thoroughly on the first two, and retain the second one for ARC-AGI submission. After identifying the strengths and weaknesses of both of these approaches, we suggest the third as a potential solution, and run preliminary experiments.