AdapterSwap: Continuous Training of LLMs with Data Removal and Access-Control Guarantees

Large language models (LLMs) are increasingly capable of completing knowledge intensive tasks by recalling information from a static pretraining corpus. Here we are concerned with LLMs in the context of evolving data requirements. For instance: batches of new data that are introduced periodically; subsets of data with user-based access controls; or requirements on dynamic removal of documents with guarantees that associated knowledge cannot be recalled. We wish to satisfy these requirements while at the same time ensuring a model does not forget old information when new data becomes available. To address these issues, we introduce AdapterSwap, a training and inference scheme that organizes knowledge from a data collection into a set of low-rank adapters, which are dynamically composed during inference. Our experiments demonstrate AdapterSwap's ability to support efficient continual learning, while also enabling organizations to have fine-grained control over data access and deletion.

@article{fleshman2025_2404.08417,
  title={ AdapterSwap: Continuous Training of LLMs with Data Removal and Access-Control Guarantees },
  author={ William Fleshman and Aleem Khan and Marc Marone and Benjamin Van Durme },
  journal={arXiv preprint arXiv:2404.08417},
  year={ 2025 }
}