#artificial intelligence
May 2026
ICICLE: Expanding Retrieval with In-Context Documents
This work revisits incremental GR as an in-context retrieval problem, where newly added documents are supplied as inference-time document-docid evidence and proposes ICICLE, an in-context indexing framework that performs source-aware docid generation over both parametric memory and context-provided document-docid pairs.
Yuliia Den, Yung-Yu Shih, Zhi Rui Tam et al.
· arXiv.org · 0 citations