Resolving Ambiguity in Composed Image Retrieval via Calibrated Interaction

Amsisan Tran, Baogh Le, Tuan Kiet Pham, Sui Yang Guang · 2026

Read Resolving Ambiguity in Composed Image Retrieval via Calibrated Interaction by Amsisan Tran, Baogh Le, Tuan Kiet Pham, Sui Yang Guang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-24634.txt.

Description

Amsisan Tran and others rethink image retrieval that pairs a photo with a text edit. A query like make it more formal points to a region of images, not one answer. Scoring such queries against a single target thus punishes good matches. Their fix gives back a candidate shortlist with a promised hit rate, and the size of that shortlist gauges query ambiguity. When the list stays large, it asks one useful follow-up question. Tests show it reaches the target with far less back-and-forth.

Catalog record

CREATORAmsisan Tran, Baogh Le, Tuan Kiet Pham, Sui Yang Guang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE42728
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.24634

Browse the full library · About the Tower of Babel