论文标题
不自然的说明：使用（几乎）没有人工劳动的语言模型
Unnatural Instructions: Tuning Language Models with (Almost) No Human Labor
论文作者
论文摘要
指令调整使经过预处理的语言模型能够从推理时间自然语言描述中执行新任务。这些方法以众包数据集或用户交互的形式依赖大量的人类监督。在这项工作中，我们介绍了不自然的说明：一大批创意和多样的指示，几乎没有人类劳动。我们通过提示一个语言模型，其中包含三个种子示例，并引起第四个种子示例。然后，通过提示该模型重新绘制每个指令，从而扩展该集合，从而总共创建约240,000个指令，输入和输出示例。实验表明，尽管包含大量的噪音，但对不自然指令进行培训与开源手动策划数据集进行训练的有效性，超过了跨各种基准测试的模型的性能，例如T0 ++和TK-Instruct。这些结果证明了模型生成数据的潜力，作为用于数据集扩展和多样化的众包众包的一种成本效益的替代方法。
Instruction tuning enables pretrained language models to perform new tasks from inference-time natural language descriptions. These approaches rely on vast amounts of human supervision in the form of crowdsourced datasets or user interactions. In this work, we introduce Unnatural Instructions: a large dataset of creative and diverse instructions, collected with virtually no human labor. We collect 64,000 examples by prompting a language model with three seed examples of instructions and eliciting a fourth. This set is then expanded by prompting the model to rephrase each instruction, creating a total of approximately 240,000 examples of instructions, inputs, and outputs. Experiments show that despite containing a fair amount of noise, training on Unnatural Instructions rivals the effectiveness of training on open-source manually-curated datasets, surpassing the performance of models such as T0++ and Tk-Instruct across various benchmarks. These results demonstrate the potential of model-generated data as a cost-effective alternative to crowdsourcing for dataset expansion and diversification.
