# ICICLE: Expanding Retrieval with In-Context Documents

Source: [ICICLE: Expanding Retrieval with In-Context Documents](<https://www.sota2.com/research/paper/6a167ac3dbd7150506f80166>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-29T05:00:05.519Z

Authors: Yu-Chen Den, Yung-Yu Shih, Zhi Rui Tam, Kuan-Yu Chen, Pu-Jen Cheng, Yun-Nung Chen, Eugene Yang

Published: 2026-05-26T12:01:44.000Z

## Abstract

Generative retrieval (GR) maps queries directly to document identifiers (docids) using parametric knowledge, However, this design makes corpus expansion costly: adding new documents requires updating model parameters to encode new document-docid associations incurs repeated training and catastrophic forgetting of previously indexed documents. In this work, we revisit incremental GR as an in-context retrieval problem, where newly added documents are supplied as inference-time document-docid evidence. We propose ICICLE, an in-context indexing framework that performs source-aware docid generation over both parametric memory and context-provided document-docid pairs. ICICLE combines a \`\[COPY\]\`-based routing mechanism, preference-based calibration, and large context adaptation to distinguish context-grounded retrieval from parametric retrieval. Experiments on MS MARCO and NQ320K show that ICICLE improves retrieval of newly introduced documents while preserving seen-document retention without corpus-specific retraining. Our analysis further shows that high-shot degradation is mainly caused by routing failure, highlighting source-selection calibration as a key bottleneck for scaling in-context generative retrieval.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Dynamic Corpus Retrieval on NQ320K D (train)](<https://www.sota2.com/research/sota/dynamic-corpus-retrieval-on-nq320k-d-train.md>): Dynamic Corpus Retrieval; NQ320K D (train); Hits@1: 0.483.

- [Dynamic Corpus Retrieval on NQ320K D\_new](<https://www.sota2.com/research/sota/dynamic-corpus-retrieval-on-nq320k-d_new.md>): Dynamic Corpus Retrieval; NQ320K D\_new; Hits@1: 64.9.

- [Dynamic Corpus Retrieval on MS MARCO D\_new](<https://www.sota2.com/research/sota/dynamic-corpus-retrieval-on-ms-marco-d_new.md>): Dynamic Corpus Retrieval; MS MARCO D\_new; Hits@1: 60.7.

- [Dynamic Corpus Retrieval on MS MARCO D (train)](<https://www.sota2.com/research/sota/dynamic-corpus-retrieval-on-ms-marco-d-train.md>): Dynamic Corpus Retrieval; MS MARCO D (train); Hits@1: 51.9.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.26902>)

[PDF](<https://arxiv.org/pdf/2605.26902.pdf>)