Multi-Modal Experience Inspired AI Creation

人工智能的创作，例如诗歌或歌词的产生，已经引起了业界和学术界的越来越多的关注，在过去几年中提出了许多有前途的模型。现有方法通常基于单个和独立的视觉或文本信息来估计输出。然而，在现实中，人类通常根据他们的经验进行创作，这可能涉及不同的模态，并且是顺序相关的。

AI-driven creation, such as the generation of poetry or lyrics, has garnered increasing attention from both industry and academia, with numerous promising models proposed in recent years. Existing approaches typically estimate outputs based on single, isolated visual or textual information. However, in reality, humans often create based on their experiences, which may involve multiple modalities and are sequence-dependent.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于多模态经验启发的AI创作，如诗歌或歌词生成，旨在克服现有方法依赖单模态信息的局限，模拟人类基于多模态、顺序相关经验的创作过程。由中国人民大学与华为于2022年联合发布，提供GitHub代码和学术论文作为参考资源。