• 2024-12-02Language models scale reliably with over-training and on downstream tasks
    本文是LLM系列文章,针对《Languagemodelsscalereliablywithover-trainingandondownstreamtasks》的翻译。语言模型可以通过过度训练和下游任务可靠地扩展摘要1引言2为过度训练和下游任务制定缩放法3构建缩放试验台4结果:可靠的推断5相关工作6局限性