# World Model on Million-Length Video andLanguage with RingAttention

Source: https://largeworldmodel.github.io
Kiwi News: https://news.kiwistand.com/stories/World-Model-on-Million-Length-Video-andLanguage-with-RingAttention?index=0x65ce77f0e0e3007f213bfdce6de7337d6aae271f9075554fd7c0cbede8dc4ee23d5e4e2e
Submitted by ruycer.eth on 2024-02-15 UTC, 2 upvotes, 1 comment

## Summary

The LargeWorldModel Organization lists two research projects on models that handle very long multimodal sequences. World Model on Million-Length Video and Language With Blockwise RingAttention, by Hao Liu, Wilson Yan, Matei Zaharia and Pieter Abbeel, is a vision-language model trained on million-length sequences of images, language and video. ElasticTok, by Wilson Yan, Matei Zaharia, Volodymyr Mnih, Pieter Abbeel, Aleksandra Faust and Hao Liu, is an adaptive tokenization model that encodes images and long video into variable-length sequences.

## Article

LargeWorldModel Organization

Below provides an overview of research related the LargeWorldModel Org.

Research

ElasticTok: Adaptive Tokenization for Image and Video Wilson Yan*, Matei Zaharia, Volodynmyr Mnih, Pieter Abbeel, Aleksandra Faust, Hao Liu*

project page tweet

An adaptive tokenization model that can encode image and long video to variable-length sequence.

World Model on Million-Length Video and Language With Blockwise RingAttention Hao Liu*, Wilson Yan*, Matei Zaharia, Pieter Abbeel

project page tweet

A vision-language model trained on million-length sequences of images, language, and video.

Website Source

## Comments

**ruycer.eth**: This paper was submitted in the AI channel on forecaster. I find it fascinating, as it has a potential to learn from video, and that opens up the possibility to learn from an endless source of good data. You can imagine learning from video from the world, and making a model of it.