Shanghao Shi, Jian Xiao, Hao Li, Yuanhaur Chang, Yevgeniy Vorobeychik, Nathan Jacobs, Neal Patwari, Ning Zhang · 2026 · arXiv
DOI: 10.48550/arXiv.2610.30439 · open-access PDF
최근의 대규모 추론 모델은 평범한 사진이 단지 시각적 정보를 사용하여 촬영된 곳을 추론할 수 있으며 메타데이터와 명백한 지리 식별자가 제거되었을 때조차도 심각한 위치 개인 정보 위험을 야기합니다. 그러한 미묘한 지역화의 메커니즘을 이해하려면 GeoTrace를 개발했으며, 모델이 위치 관련 시각적 힌트를 추출하고 거친 지리적 추론을 통해 그들을 구성하는 방법을 보여주는 구조화 된 추론 공격입니다.
Machine translation (Meta M2M100); the English original follows.
Recent large reasoning models can infer where ordinary photos were taken using only visual information, creating a serious location privacy risk even when metadata and obvious geographic identifiers are removed. To understand the mechanism behind such fine-grained localization, we develop GeoTrace, a structured reasoning attack that exposes how models extract location-relevant visual clues and compose them through coarse-to-fine geographic reasoning. We evaluate GeoTrace on a worldwide benchmark of 525 ordinary images spanning natural, urban, residential, and human-centered scenes while exclud
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.30439, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange