arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.13372cs.AIcs.LG
收藏

为AGI进步的ARC:抽象与推理的活体调查

The ARC of Progress towards AGI: A Living Survey of Abstraction and Reasoning

TIB - Leibniz Information Centre for Science and Technology(TIB - 科学技术信息研究中心);L3S Research Center, Leibniz University of Hannover(L3S研究所以汉诺威莱布尼茨大学);Dresden University of Technology, Amazon (work done outside of Amazon)(德累斯顿技术大学,亚马逊(非亚马逊工作))

Sahar Vahdati, Andrei Aioanei, Haridhra Suresh, Jens Lehmann

AI总结:

本文通过分析三个版本的82种方法及ARC Prize竞赛,揭示了不同范式在版本间性能下降的规律,指出组合泛化能力的局限性,并强调测试时适应和精炼循环的重要性。

摘要:

抽象与推理语料库(ARC-AGI)已成为AI流体智力的关键基准。本文首次对82种方法在三个基准版本及ARC Prize 2024-2025竞赛中的跨世代分析表明,所有范式在版本间性能下降一致:程序合成、神经符号和神经方法在ARC-AGI-1到ARC-AGI-2间下降2-3倍,表明组合泛化能力的根本限制。尽管系统在ARC-AGI-1上达到93.0%(Opus 4.6),但在ARC-AGI-2和ARC-AGI-3上分别降至68.8%和13%,而人类在所有版本中保持几乎完美准确度。成本在一年内下降了390倍(o3的$4,500/任务到GPT-5.2的$12/任务),尽管这主要反映的是测试时并行度的减少。万亿级模型在得分和成本上差异很大,而Kaggle受限的条目(660M-8B)实现了具有竞争力的结果,符合Chollet关于智能是技能获取效率的论点。测试时适应和精炼循环成为关键成功因素,而组合推理和交互学习仍待解决。ARC Prize 2025的获胜者需要数万例合成示例才能在ARC-AGI-2上达到24%,证实推理仍受知识限制。本ARC-AGI活体调查的首次发布截至2026年2月,更新可在https://nimi-ai.com/arc-survey/

英文摘要:

The Abstraction and Reasoning Corpus (ARC-AGI) has become a key benchmark for fluid intelligence in AI. This survey presents the first cross-generation analysis of 82 approaches across three benchmark versions and the ARC Prize 2024-2025 competitions. Our central finding is that performance degradation across versions is consistent across all paradigms: program synthesis, neuro-symbolic, and neural approaches all exhibit 2-3x drops from ARC-AGI-1 to ARC-AGI-2, indicating fundamental limitations in compositional generalization. While systems now reach 93.0% on ARC-AGI-1 (Opus 4.6), performance falls to 68.8% on ARC-AGI-2 and 13% on ARC-AGI-3, as humans maintain near-perfect accuracy across all versions. Cost fell 390x in one year (o3's $4,500/task to GPT-5.2's $12/task), although this largely reflects reduced test-time parallelism. Trillion-scale models vary widely in score and cost, while Kaggle-constrained entries (660M-8B) achieve competitive results, aligning with Chollet's thesis that intelligence is skill-acquisition efficiency. Test-time adaptation and refinement loops emerge as critical success factors, while compositional reasoning and interactive learning remain unsolved. ARC Prize 2025 winners needed hundreds of thousands of synthetic examples to reach 24% on ARC-AGI-2, confirming that reasoning remains knowledge-bound. This first release of the ARC-AGI Living Survey captures the field as of February 2026, with updates at https://nimi-ai.com/arc-survey/

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑