While model capabilities have improved dramatically for use cases like codegen and mathematical reasoning, they still lag behind on the data side (as evidenced through SQL benchmarks like Spider 2.0 and Bird Bench).
令人惊讶的是：尽管AI模型在代码生成和数学推理方面取得了巨大进步，但在数据处理方面仍然落后。Spider 2.0和Bird Bench等基准测试显示，AI在SQL查询等基础数据任务上表现不佳，这表明当前AI技术存在明显的应用局限性。