Hard Vision, Easy Vision: What GPT-6 Astra Reveals Across Computer Vision

• 类型：arxiv
• 标识：2609.35718
• 链接：https://arxiv.org/abs/2609.35718
• 主分类：multimodal
• 形态：benchmark
• TLDR：Frontier general-purpose systems are rapidly expanding beyond visual understanding into capabilities traditionally handled by dedicated computer-vision models. As these capabilities expand, a central question for the computer-vision community is how far this reach extends, and what remains hard. We evaluate GPT-6 Astra alongside five frontier general-purpose AI systems across 34 capabilities and 55 benchmarks spanning nine areas of computer vision. We compare their performance with dedicated models and humans where suitable references are available. Astra demonstrates broad visual capability,
• 副分类：evaluation
• 待LLM分类：否
• 标题中文：难视觉，易视觉：GPT-6 Astra 揭示的计算机视觉全貌
• TLDR中文：前沿通用系统正快速从视觉理解扩展到传统由专用计算机视觉模型处理的能力。随着能力扩展，计算机视觉社区的核心问题是这种延伸能走多远，以及什么仍属困难。我们在 34 项能力、55 个基准（覆盖计算机视觉 9 个领域）上评估 GPT-6 Astra 与 5 个前沿通用 AI 系统，并在有可用参考处与专用模型及人类对比。Astra 展现出广泛的视觉能力
• 来源文件：
• /inbox/tom/_candidates/2026-09-29-agent-rag-longcontext-candidates.json