跳到正文
@EpochAIResearch· @EpochAIResearch · X·· 2026-08-29AI 评分36
AI 导读

大约一半的人类得分超过了 AI,但几乎所有人都至少表现出了一定程度的随时间提升。我们评测的几个 AI 系统在首次游玩时的表现就优于几乎所有人类的首次游玩,但这些 AI 此后最多只表现出非常轻微的提升。https://t.co/Sxw7GuxZDe

正文

About half of humans outscored AI, but almost all showed at least some improvement over time. Several AI systems we benchmarked were better on their first playthrough than almost all first human playthroughs, but these AI showed at most very slight improvement from there. https://t.co/Sxw7GuxZDe

来源:@EpochAIResearch · x.com