Next Token Prediction Towards Multimodal Intelligence: A Comprehensive Survey
Abstract
Building on the foundations of language modeling in natural language processing, Next Token Prediction (NTP) has evolved into a versatile training objective for machine learning tasks across various modalities, achieving considerable success. As Large Language Models (LLMs) have advanced to unify understanding and generation tasks within the textual modality, recent research has shown that tasks from different modalities can also be effectively encapsulated within the NTP framework, transforming the multimodal information into tokens and predict the next one given the context. This survey introduces a comprehensive taxonomy that unifies both understanding and generation within multimodal learning through the lens of NTP. The proposed taxonomy covers five key aspects: Multimodal tokenization, MMNTP model architectures, unified task representation, datasets \& evaluation, and open challenges. This new taxonomy aims to aid researchers in their exploration of multimodal intelligence. An associated GitHub repository collecting the latest papers and repos is available at https://github.com/LMM101/Awesome-Multimodal-Next-Token-Prediction
Cite
@article{arxiv.2412.18619,
title = {Next Token Prediction Towards Multimodal Intelligence: A Comprehensive Survey},
author = {Liang Chen and Zekun Wang and Shuhuai Ren and Lei Li and Haozhe Zhao and Yunshui Li and Zefan Cai and Hongcheng Guo and Lei Zhang and Yizhe Xiong and Yichi Zhang and Ruoyu Wu and Qingxiu Dong and Ge Zhang and Jian Yang and Lingwei Meng and Shujie Hu and Yulong Chen and Junyang Lin and Shuai Bai and Andreas Vlachos and Xu Tan and Minjia Zhang and Wen Xiao and Aaron Yee and Tianyu Liu and Baobao Chang},
journal= {arXiv preprint arXiv:2412.18619},
year = {2024}
}
Comments
69 papes, 18 figures, repo at https://github.com/LMM101/Awesome-Multimodal-Next-Token-Prediction