Import AI 476：创造力与递归自我改进；开源权重模型表现逊于中国；AI 走向太空
原标题：Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space
AI 摘要
AI 初创公司 Elorian 与 Scale AI 发布 Humanity's Sixth Sense (HSS) 基准，用 522 道题、288 张图和 234 段视频测试多模态模型的直觉视觉推理能力。人类基线 pass@1 达 93.1，而最佳 AI 系统 GPT 6 Astra 仅 53.6，中国开源权重模型 GLM-5.3-flash 和 Kimi K3 分别只有 26.8 和 25.5。文章还讨论了 AI 推动太空经济与自主系统、以及 P-Zero Research 的 TasteVal 基准测试 AI 研究品味与递归自我改进潜力。
正文节选
Welcome to Import AI, a newsletter about AI research. Import AI runs on arXiv, cappuccinos, and feedback from readers. If you’d like to support this, please subscribe. New benchmark shows how AI models lack intuitive visual reasoning: …Humanity’s Sixth Sense highlights some shortcomings of current AI systems… AI startups Elorian and Scale AI have built and released Humanity’s Sixth Sense, a benchmark that “measures how a model interprets and reasons through information a visual scene implies but