Built independently by an author, for readers. Read the story and support ChapterPal

keyword

unstructured identifiers

Unstructured identifiers are discrete, arbitrary labels assigned to documents in generative information retrieval systems that carry no inherent semantic or hierarchical meaning. In models that directly map search queries to document representations, an unstructured identifier treats each document as an independent, atomic unit, typically encoded as a single unique token or arbitrary integer. Under this approach, the retrieval process is framed as a classification task over the entire set of individual document tokens, contrasting with structured identifier methods that utilize hierarchical clusters, semantic strings, or sequential subword representations to reflect relationships among documents.

1 item

Transformer Memory as a Differentiable Search Index

Transformer Memory as a Differentiable Search Index

Yi Tay, Vinh Tran, Mostafa Dehghani, Jianmo Ni, Dara Bahri, Harsh Mehta, Zhen Qin, Kai Hui, Zhe Zhao, Jai Prakash Gupta, Tal Schuster, William W. Cohen, Donald Metzler

OrganizationsGoogle

Why you should read this

Proposes the Differentiable Search Index, demonstrating that a single Transformer can internalize an entire text corpus within its parameters to map queries directly to document identifiers, outperforming standard dual-encoder retrieval systems.

In this paper, we demonstrate that information retrieval can be accomplished with a single Transformer, in which all information about the corpus is encoded in the parameters of the model. To this end, we introduce the Differentiable Search Index (DSI), a new paradigm that learns a text-to-text model that maps string queries directly to relevant docids; in other words, a DSI model answers queries directly using only its parameters, dramatically simplifying the whole retrieval process. We study variations in how documents and their identifiers are represented, variations in training procedures, and the interplay between models and corpus sizes. Experiments demonstrate that given appropriate design choices, DSI significantly outperforms strong baselines such as dual encoder models. Moreover, DSI demonstrates strong generalization capabilities, outperforming a BM25 baseline in a zero-shot setup.

Added

2026-09-26