ARC Prize summarizes 30 days of ARC-AGI-3 Preview and agent competition results
Original title:ARC-AGI-3 Preview: 30-Day Learnings
AI summary
The ARC Prize Foundation summarized 30 days of ARC-AGI-3 Preview data: more than 1,200 people played over 3,900 games, humans mostly completed them, and AI agents made inefficient progress. Competition winner StochasticGoose scored 12.58% and completed 18 levels, with the top three agents all based on intelligent random exploration; officials noted some games can be solved by brute force and plan to score against human baselines by action efficiency and add improvements such as an undo button and offline engine.
Key Takeaways
- The ARC Prize Foundation summarized 30 days of ARC-AGI-3 Preview data: more than 1,200 people…
- Competition winner StochasticGoose scored 12.
- 58% and completed 18 levels,with the top three agents all based on intelligent random exploration
- officials noted some games can be solved by brute force and plan to score against human baselines…
Quick Facts
- Published
- Updated
- Source
- ARC Prize:官方博客
- Category
- Papers & Research
- AI score
- 61
原文公布了30天竞赛的完整人类与Agent得分对比,读者可以看到交互式推理基准用动作效率区分人类和AI的具体依据。
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 summarizes 30 days of ARC-AGI-3 Preview and agent competition results. August 19, 2025. https://new.nat6.net/en/news/cmtyo2p2s03g1rog04vw1a1ch/