Infinite-parameter Large Language Model
Keywords: lifelong learning
TL;DR: Achieving a large model with a dynamic number of parameters through routing.
Abstract: In the standard transformer architecture, increasing model parameters leads to linear growth in computational cost and activation memory. To address this issue, we propose a novel Infinite Parameter Large Language Model (IP-LLM) architecture that decouples model size from computational cost and device memory. Existing large language models are all fixed-parameter models, while human knowledge is infinite and expands daily. Finite parameters are inherently limited in their capacity to accommodate this boundless knowledge. Our IP-LLM architecture can potentially accommodate infinite knowledge, resolving this issue and laying the foundation for realizing a truly omniscient and omnipotent artificial general intelligence in the future.
Primary Area: transfer learning, meta learning, and lifelong learning
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.
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: 6013
Loading