Large Dual Encoders Are Generalizable Retrievers

It has been shown that dual encoders trained on one domain often fail to generalize to other domains for retrieval tasks. One widespread belief is that the bottleneck layer of a dual encoder, where the final score is simply a dotproduct between a query vector and a passage vector, is too limited to make dual encoders an effective retrieval model for out-ofdomain generalization. In this paper, we challenge this belief by scaling up the size of the dual encoder model while keeping the bottleneck embedding size fixed. With multi-stage training, surprisingly, scaling up the model size brings significant improvement on a variety of retrieval tasks, especially for out-ofdomain generalization. Experimental results show that our dual encoders, Generalizable T5-based dense Retrievers (GTR), outperform existing sparse and dense retrievers on the BEIR dataset (Thakur et al., 2021) significantly. Most surprisingly, our ablation study finds that GTR is very data efficient, as it only needs 10% of MS Marco supervised data to achieve the best out-of-domain performance. 1