ARC Prize releases ARC-AGI-2 technical report: frontier models score under 5%, humans solve 100% of tasks
Original title:ARC-AGI-2 A New Challenge for Frontier AI Reasoning Systems
AI summary
ARC Prize released the ARC-AGI-2 technical report, introducing a new benchmark for frontier AI reasoning systems with tasks designed to remain easy for humans and hard for AI while reducing the feasibility of brute force. Testing with 400 people covered 1,417 tasks, all of which humans could solve, averaging about 2.3 minutes per task; at release, leading models all scored below 5% on ARC-AGI-2, whereas comparable models typically scored 20% to 50% on ARC-AGI-1.
Key Takeaways
- ARC Prize released the ARC-AGI-2 technical report,introducing a new benchmark for frontier AI…
- Testing with 400 people covered 1,417 tasks,all of which humans could solve,averaging about 2.
- 3 minutes per task
- at release,leading models all scored below 5% on ARC-AGI-2,whereas comparable models typically…
Quick Facts
- Published
- Updated
- Source
- ARC Prize:官方博客
- Category
- Papers & Research
- AI score
- 67
官方技术报告给出人类实测与前沿模型得分对比,可以据此了解新基准的难度设计和它想度量的能力差距。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: ARC Prize releases ARC-AGI-2 technical report: frontier models score under 5%, humans solve 100% of tasks. May 20, 2025. https://new.nat6.net/en/news/cmtyo2p2s03g5rog00pkk4enc/