Making Dialogue Grounding Data Rich: A Three-Tier Data Synthesis Framework for Generalized Referring Expression Comprehension

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shao, Juexi, Li, Siyou, Gan, Yujian, Madge, Chris, Karan, Vanja, Poesio, Massimo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911622194266112
author Shao, Juexi
Li, Siyou
Gan, Yujian
Madge, Chris
Karan, Vanja
Poesio, Massimo
author_facet Shao, Juexi
Li, Siyou
Gan, Yujian
Madge, Chris
Karan, Vanja
Poesio, Massimo
contents Dialogue-Based Generalized Referring Expression Comprehension (GREC) requires models to ground the expression and unlimited targets in complex visual scenes while resolving coreference across a long dialogue context. However, existing systems struggle under distribution shift between training and evaluation domains, a gap exacerbated by the scarcity of annotated dialogue grounding data. We address this challenge with a three-tier data-synthesis method that balances realism and controllability to produce scalable supervision for dialogue-conditioned grounding. Fine-tuning on the synthesized data yields consistent, substantial improvements over prior approaches across standard evaluation metrics.
format Preprint
id arxiv_https___arxiv_org_abs_2512_02791
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Making Dialogue Grounding Data Rich: A Three-Tier Data Synthesis Framework for Generalized Referring Expression Comprehension
Shao, Juexi
Li, Siyou
Gan, Yujian
Madge, Chris
Karan, Vanja
Poesio, Massimo
Computation and Language
Dialogue-Based Generalized Referring Expression Comprehension (GREC) requires models to ground the expression and unlimited targets in complex visual scenes while resolving coreference across a long dialogue context. However, existing systems struggle under distribution shift between training and evaluation domains, a gap exacerbated by the scarcity of annotated dialogue grounding data. We address this challenge with a three-tier data-synthesis method that balances realism and controllability to produce scalable supervision for dialogue-conditioned grounding. Fine-tuning on the synthesized data yields consistent, substantial improvements over prior approaches across standard evaluation metrics.
title Making Dialogue Grounding Data Rich: A Three-Tier Data Synthesis Framework for Generalized Referring Expression Comprehension
topic Computation and Language
url https://arxiv.org/abs/2512.02791