AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

GeoBiaset: A Counterfactual Benchmark for Demographic Bias in World-Level Geolocalization

Niccolò Niccoli, Silvia Dani, Marco Mistretta, Andrew Bagdanov, Marco Bertini, Lorenzo Seidenari

NeurIPS 2026 · Image generation › Autoregressive and latent generation · approach: Retrieval and matching with generative features · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Introduces GeoBiaset, a counterfactual benchmark for auditing demographic bias in world-level image geolocalization by examining which visual cues drive regional disparities.

Problem

Current world-level image geolocalization models show biases (e.g., missing cities in training data, better performance in wealthier regions), but it is unclear which visual cues drive these disparities.

Approach

A counterfactual benchmark (GeoBiaset) that is meant to isolate visual cues responsible for geolocalization disparities across demographic/regional groups; the abstract is truncated so the exact construction is not stated.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
insert a person into the background imageGeoCLIP · GeoBiasetcontinent accuracy (%): 81.9 -> 55.3Hurt
insert a person into the background imageImg2Loc · GeoBiasetcontinent accuracy (%): 70.3 -> 37.5Hurt
insert a person into the background imageG3 · GeoBiasetcontinent accuracy (%): 81.6 -> 47.8Hurt
insert a person into the background imageGLM-4.1V (zero-shot MLLM) · GeoBiasetcontinent accuracy (%): 82.8 -> 69.2Hurt
mask the inserted person (black) on both original and edited images (ablation)GeoCLIP · GeoBiasetcontinent accuracy (%): 55.3 -> 72.7Improved
mask the inserted person (gray) on both original and edited images (ablation)GeoCLIP · GeoBiasetcontinent accuracy (%): 55.3 -> 71.7Improved
blur faces of inserted person (ablation)GeoCLIP · GeoBiasetaccuracy at 2500km: +8.8 (Black), +11.7 (Latino Hispanic), +1.7 (White)Improved
use race-associated attire for inserted person (ablation)GeoCLIP · GeoBiaset, Black subjectcontinent accuracy change (%): -15.5Hurt
use race-associated attire for inserted person (ablation)GeoCLIP · GeoBiaset, Middle Eastern subjectcontinent accuracy change (%): +0.4No clear effect
scene contains a landmark vs not (ablation)GeoCLIP · GeoBiasetPOPI-H accuracy change (%): 27.1 -> 9.2Improved
scene contains a landmark vs not (ablation)GeoCLIP · GeoBiasetPOPI-D accuracy change (%): 33.2 -> 26.0Improved
use dedicated geolocalization model instead of zero-shot MLLMgeolocalization models · GeoBiasetrelative continent drop: 33-47% vs 16-25%Hurt

Limitations

Experimental setup

Benchmarks: GeoBiaset

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy