Merlin:Empowering Multimodal LLMs with Foresight Minds
2024-07-03
状态已发表
摘要

Humans can foresee the future based on present observations, a skill we term as foresight minds. . However, this capability remains under-explored within existing MLLMs, hindering their capacity to understand intentions behind subjects. To address this, we integrate the future modeling into MLLMs. By utilizing the trajectory, , a highly structured representation, as a learning objective, we aim to equip the model to understand spatiotemporal dynamics. Inspired by the learning paradigm of LLMs, we first propose Foresight Pre-Training (FPT) that jointly learns various tasks centered on trajectories, enabling MLLMs to predict entire trajectories from a given initial observation. Then, we propose Foresight Instruction-Tuning (FIT) that requires MLLMs to reason about potential future events based on predicted trajectories. Aided by FPT and FIT, we build an unified MLLM named Merlin that supports complex future reasoning. Experiments show Merlin’s foresight minds with impressive performance on both future reasoning and visual comprehension tasks. 

关键词Multimodal Large Language Model Future Reasoning
DOIarXiv:2312.00589
相关网址查看原文
出处Arxiv
WOS记录号PPRN:86357261
WOS类目Computer Science, Software Engineering
文献类型预印本
条目标识符https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/404293
专题信息科学与技术学院_硕士生
通讯作者Tao, Wenbing
作者单位
1.Huazhong Univ Sci & Technol, Wuhan, Peoples R China
2.MEGVII Technol, Beijing, Peoples R China
3.ShanghaiTech Univ, Shanghai, Peoples R China
4.Beijing Inst Technol, Beijing, Peoples R China
5.Univ Chinese Acad Sci, Beijing, Peoples R China
推荐引用方式
GB/T 7714
Yu, En,Zhao, Liang,Wei, Yana,et al. Merlin:Empowering Multimodal LLMs with Foresight Minds. 2024.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
个性服务
查看访问统计
谷歌学术
谷歌学术中相似的文章
[Yu, En]的文章
[Zhao, Liang]的文章
[Wei, Yana]的文章
百度学术
百度学术中相似的文章
[Yu, En]的文章
[Zhao, Liang]的文章
[Wei, Yana]的文章
必应学术
必应学术中相似的文章
[Yu, En]的文章
[Zhao, Liang]的文章
[Wei, Yana]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。