📄 中文摘要
2026年1月的评估结果显示，通用模型结合工具的表现超越了专门化的设置。这项研究分析了48000个答案，揭示了工具增强的AI在多个科学任务中的优势。研究表明，通用模型不仅在准确性上有所提升，同时在处理复杂问题时展现出更强的适应能力。这一发现为未来AI技术的发展提供了新的方向，强调了工具与模型结合的重要性。
OpenAI评估48000个答案：工具增强的AI在16项科学任务中表现优异
出处: OpenAI Reviews 48000 Answers Tool-Augmented AI Tops 16 Scientific Tasks
发布: 2026年3月29日
📄 English Summary
OpenAI Reviews 48000 Answers Tool-Augmented AI Tops 16 Scientific Tasks
The evaluation conducted in January 2026 revealed that general-purpose models augmented with tools outperformed specialized setups across 16 scientific tasks. This study analyzed 48,000 answers, highlighting the advantages of tool-augmented AI in various scientific challenges. The findings indicate that general-purpose models not only improved in accuracy but also demonstrated greater adaptability in handling complex problems. This discovery offers new directions for the future development of AI technologies, emphasizing the significance of integrating tools with models.