ED2LM: Encoder-Decoder to Language Model for Faster Document Re-ranking Inference
This work addresses efficiency issues in neural re-ranking for information retrieval, offering a faster alternative to existing models, though it is incremental as it builds on pretrained encoder-decoder architectures.
The paper tackles the high computational cost of cross-attention models in document re-ranking by proposing a new training and inference paradigm that finetunes an encoder-decoder model for document-to-query generation and decomposes it into a decoder-only language model at inference, achieving results comparable to existing methods while being up to 6.8 times faster.
State-of-the-art neural models typically encode document-query pairs using cross-attention for re-ranking. To this end, models generally utilize an encoder-only (like BERT) paradigm or an encoder-decoder (like T5) approach. These paradigms, however, are not without flaws, i.e., running the model on all query-document pairs at inference-time incurs a significant computational cost. This paper proposes a new training and inference paradigm for re-ranking. We propose to finetune a pretrained encoder-decoder model using in the form of document to query generation. Subsequently, we show that this encoder-decoder architecture can be decomposed into a decoder-only language model during inference. This results in significant inference time speedups since the decoder-only architecture only needs to learn to interpret static encoder embeddings during inference. Our experiments show that this new paradigm achieves results that are comparable to the more expensive cross-attention ranking approaches while being up to 6.8X faster. We believe this work paves the way for more efficient neural rankers that leverage large pretrained models.