🤖 AI Summary
Existing generative retrieval (GR) methods rely on fine-tuning models to construct trainable indexes, suffering from high training costs, underutilization of pretrained knowledge, and poor adaptability to dynamic document collections. To address these limitations, we propose a few-shot index-based generative retrieval framework that eliminates model fine-tuning: instead, it prompts large language models (LLMs) with a small set of demonstrations to construct an “ID bank” over the document collection; at inference, the LLM is constrained to generate only IDs from this bank. We introduce the first few-shot indexing paradigm and a one-to-many mapping mechanism to enhance ID generation robustness, thereby fully harnessing LLMs’ inherent retrieval capabilities. Experiments demonstrate that our method surpasses state-of-the-art GR approaches requiring extensive training across multiple benchmarks, while drastically reducing computational overhead and enabling rapid adaptation to evolving corpora.
📝 Abstract
Existing generative retrieval (GR) approaches rely on training-based indexing, i.e., fine-tuning a model to memorise the associations between a query and the document identifier (docid) of a relevant document. Training-based indexing has three limitations: high training overhead, under-utilization of the pre-trained knowledge of large language models (LLMs), and challenges in adapting to a dynamic document corpus. To address the above issues, we propose a novel few-shot indexing-based GR framework (Few-Shot GR). It has a novel few-shot indexing process, where we prompt an LLM to generate docids for all documents in a corpus, ultimately creating a docid bank for the entire corpus. During retrieval, we feed a query to the same LLM and constrain it to generate a docid within the docid bank created during indexing, and then map the generated docid back to its corresponding document. Few-Shot GR relies solely on prompting an LLM without requiring any training, making it more efficient. Moreover, we devise few-shot indexing with one-to-many mapping to further enhance Few-Shot GR. Experiments show that Few-Shot GR achieves superior performance to state-of-the-art GR methods that require heavy training.