Unbounded cache model for online language modeling with open vocabulary
This work addresses the challenge of distribution shift in online language modeling for applications requiring adaptation to evolving data, though it is incremental as it extends existing cache models.
The paper tackled the problem of adapting pre-trained language models to new data distributions by scaling continuous cache models to larger contexts, resulting in significant perplexity improvements on new distributions and efficient scaling to much larger contexts than previous local cache models.
Recently, continuous cache models were proposed as extensions to recurrent neural network language models, to adapt their predictions to local changes in the data distribution. These models only capture the local context, of up to a few thousands tokens. In this paper, we propose an extension of continuous cache models, which can scale to larger contexts. In particular, we use a large scale non-parametric memory component that stores all the hidden activations seen in the past. We leverage recent advances in approximate nearest neighbor search and quantization algorithms to store millions of representations while searching them efficiently. We conduct extensive experiments showing that our approach significantly improves the perplexity of pre-trained language models on new distributions, and can scale efficiently to much larger contexts than previously proposed local cache models.