FIRST: Faster Improved Listwise Reranking with Single Token Decoding

TL;DR

The paper introduces FIRST, a more efficient listwise LLM reranking method that speeds up inference by 50% and improves ranking accuracy, demonstrating practical benefits in retrieval tasks and relevance feedback.

Contribution

FIRST leverages the first generated token logits for faster reranking and incorporates a learning-to-rank loss, advancing listwise LLM reranking efficiency and effectiveness.

Findings

FIRST accelerates inference by 50%.

Maintains robust ranking performance across benchmarks.

Improves retriever recall with relevance feedback.

Abstract

Large Language Models (LLMs) have significantly advanced the field of information retrieval, particularly for reranking. Listwise LLM rerankers have showcased superior performance and generalizability compared to existing supervised approaches. However, conventional listwise LLM reranking methods lack efficiency as they provide ranking output in the form of a generated ordered sequence of candidate passage identifiers. Further, they are trained with the typical language modeling objective, which treats all ranking errors uniformly--potentially at the cost of misranking highly relevant passages. Addressing these limitations, we introduce FIRST, a novel listwise LLM reranking approach leveraging the output logits of the first generated identifier to directly obtain a ranked ordering of the candidates. Further, we incorporate a learning-to-rank loss during training, prioritizing ranking accuracy for the more relevant passages. Empirical results demonstrate that FIRST accelerates inference by 50% while maintaining a robust ranking performance with gains across the BEIR benchmark. Finally, to illustrate the practical effectiveness of listwise LLM rerankers, we investigate their application in providing relevance feedback for retrievers during inference. Our results show that LLM rerankers can provide a stronger distillation signal compared to cross-encoders, yielding substantial improvements in retriever recall after relevance feedback.