World Model on Million-Length Video And Language With Blockwise RingAttention
Keywords: long context, language model
Abstract: Enabling long-context understanding remains a key challenge in scaling existing sequence models -- a crucial component in developing generally intelligent models that can process and operate over long temporal horizons that potentially consist of millions of tokens. In this paper, we seek to make strides towards addressing these challenges by providing a comprehensive exploration into the full development process to produce 1M context language models and video-language models, setting new benchmarks in language retrieval and new capabilities in long video understanding. Furthermore, we provide details in our long context data curation process, progressive context extension from 4K to 1M tokens, and an efficient Fused Blockwise RingAttention implementation to scalably train on long sequences. As a benefit to the community, we additionally fully open-source a family of 7B parameter models capable of processing long text documents (LWM-Text, LWM-Text-Chat) and videos (LWM, LWM-Chat) of over 1M tokens.
Supplementary Material: zip
Primary Area: generative models
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11124
Loading