Event ·Multi-source

GPT-6 Astra 展示符号世界建模能力

GPT-6 Astra 在 ARC-AGI-3 基准测试中得分为 66%，使用连续对话测试时接近 100%。该模型能够高效地进行即时符号世界建模，甚至开发自己的特定领域 DSL 表示游戏情况。Astra 在行动效率上几乎超越所有级别的人类基准线，标志着模型智能的重大突破。

Current read

GPT-6 Astra 能自己创建符号世界模型，在推理任务上接近满分，比人类还高效。

11 sources73° AI heatLast updated 2026/9/3 20:50:12

Evidence

11 sources

Primary sourceGary Marcus

2026/9/3 20:50:12View source

Related source 1The Rundown AI

2026/9/3 18:07:36View source

Related source 2@OpenAIDevs

2026/9/3 19:33:24View source

Related source 3Greg Brockman

2026/9/3 19:44:36View source

Related source 4Sam Altman

2026/9/3 19:49:10View source

Related source 5orange.ai

2026/9/3 23:36:56View source

Related source 6歸藏(guizang.ai)

2026/9/4 00:59:38View source

Related source 7Latent Space (swyx)

2026/9/4 05:18:11View source

Related source 8Decoder

2026/9/4 11:07:36View source

Related source 9shao__meng

2026/9/5 02:18:40View source

Related source 10lmarena.ai

2026/9/5 17:40:41View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story