DialogueVPR: Towards Conversational Visual Place Recognition

DialogueVPR：迈向对话式视觉地点识别

Abstract: Inspired by how humans communicate spatial information, language-guided geo-localization has gained significant traction for its intuitive and practical value. Despite this progress, most methods still rely on a static, one-shot retrieval paradigm, which fails to handle the ambiguity and incompleteness inherent in real-world natural language descriptions.

摘要： 受人类交流空间信息方式的启发，语言引导的地理定位因其直观和实用的价值而备受关注。尽管取得了这些进展，但大多数方法仍然依赖于静态的“一次性”检索范式，这无法处理现实世界自然语言描述中固有的歧义性和不完整性。

We propose a paradigm shift to reasoning retrieval and introduce Dialogue Place Recognition (DlgPR), which casts localization as an interactive, dialogue-driven reasoning process. To support this new task, we present DlgQuest-Cities, the first large-scale dialogue-based benchmark for place recognition, and a unified reasoning framework that couples a cross-modal multi-level retriever with an intelligent questioner, DQ-pilot.

我们提出了一种向推理检索的范式转变，并引入了对话式地点识别（DlgPR），将定位过程转化为一种交互式的、由对话驱动的推理过程。为了支持这一新任务，我们提出了 DlgQuest-Cities，这是首个用于地点识别的大规模对话基准，以及一个统一的推理框架，该框架将跨模态多级检索器与智能提问者 DQ-pilot 相结合。

DQ-pilot is trained in a curriculum: supervised fine-tuning on a curated DQ-cities-20k subset followed by reinforcement refinement on a harder DQ-cities-10k split via GRPO. Two task-aligned metrics guide learning: a Discriminative Difficulty Index (DDI) for curriculum sampling and a Positional Retrieval Gain (PRG) reward that directly measures retrieval improvement induced by a question.

DQ-pilot 采用课程学习方式进行训练：首先在精选的 DQ-cities-20k 子集上进行监督微调，随后通过 GRPO 在更具挑战性的 DQ-cities-10k 分割集上进行强化精炼。两个与任务对齐的指标引导着学习过程：用于课程采样的判别难度指数（DDI），以及直接衡量由提问带来的检索提升的定位检索增益（PRG）奖励。

Experiments show this reasoning-based approach significantly outperforms baselines. The code and model are available at this https URL.

实验表明，这种基于推理的方法显著优于基准模型。代码和模型已在链接中提供。