Language models generate and update weights from live data streams
Infinite-Parameter LLMs: Generating and Adapting Weights from Live Data
Artificial IntelligenceMachine Learning
Summary
Language models usually learn from a fixed training dataset and cannot change once deployed. The authors propose a new model that changes its inner weights dynamically based on the live data it receives during use. This means the model can learn continuously during a conversation or interaction, storing new facts and corrections in its 'memory' rather than in the temporary prompt. Their approach uses a compact network that generates weight adjustments in real time, enabling the model to adapt without growing its stored size. This could make models more responsive and efficient compared to just using prompts to hold temporary knowledge.
What this means in practice
- •For chatbot developers: Implement chatbots that can learn from user corrections and facts during conversations, improving responsiveness without expanding model size.
- •For personal assistant platform teams: Create assistants that personalize knowledge and behavior dynamically within sessions by updating their model weights from live data.$Commercial implications: This enables selling smarter personal assistants that adapt continuously, offering better user experience with smaller computational footprints.
Authors
Jinli Hu, Ross M. Clarke, Yichuan Zhang, José Miguel Hernández-Lobato
Abstract
The scaling laws hold that a language model grows more capable with more parameters and more training data, and Mixture-of-Experts (MoE) architectures have ridden these laws to remarkable results, activating only a fraction of an enormous stored parameter bank for each token. That success is built on static pretraining data. A deployed model faces a different world, where much of the data that would make it more useful is not in its training set but in the live interaction it is currently handling, such as the facts a user supplies or the corrections they give. A conventional model cannot learn from this data, because its weights are frozen after training. Instead, the knowledge and behaviour supplied at run time are placed in the prompt, by retrieval or instruction, and re-read on every request only to be discarded once the request ends. We ask how an architecture could learn from live interaction by writing it into its weights. Taking inspiration from MoE, we propose the \textbf{Infinite-Parameter LLM}. A compact hypernetwork turns the data given at run time into a low-rank modulation of a shared base network, so the feed-forward weights are generated from live data rather than stored in a fixed bank. Where prior weight generators read the context once and freeze, we carry a Bayesian belief over the generator's latent code and update it online, so the effective weight is re-derived from that evolving belief as the session proceeds rather than fixed after one read. The stored footprint stays fixed, yet the weights the model can compile are effectively infinite. For the knowledge and behaviour supplied at run time, carrying them in the weights rather than the prompt is amortized in compute, frees the context window, persists across turns, and can generalise better than in-context use. We specify an evaluation protocol that tests exactly this against in-context learning and retrieval.