X：ARC Prize (@arcprize)· @MTSlive·· 4 天前AI 评分
Astra 在 ARC-AGI-3 上得分 99.9%，ARC Prize 推出交互式探索方式
RT by @arcprize: Astra scored 99.9% on ARC-AGI-3. What does that actually mean? We built an interactive way to explore @arcprize's ARC-AGI-3. Play one of the games that models are tested on, then read Astra's notes from its own run. https://intelligence.mts.now/benchmarks
Astra 在 ARC-AGI-3 上得分 99.9%，ARC Prize 为此构建了一个交互式页面，让用户亲自试玩模型受测的游戏，并阅读 Astra 自己运行过程留下的笔记。有评论质疑该成绩并未体现理解能力，认为关键在于弄清这些游戏的机制，而 Astra 的笔记中看不到它如何预判新关卡会发生什么。
来源：X：ARC Prize (@arcprize) · x.lingyaoai.com