AI BENCHMARK PROFILE

Gaze-Co

Health & Life SciencesMultimodal Perception

Evaluates promptable gaze target estimation: given an image and a text or visual prompt identifying a subject, models output the subject's location, in/out-of-frame status, and a gaze target heatmap.

Released: 2026-08-11
Readiness: Runnable
Primary field: Health & Life Sciences

Why it matters

Introduces concept-driven gaze estimation that removes brittle intermediate pipelines and provides reusable prompted splits across existing gaze datasets.

Motivation

Estimating human gaze targets from images in-the-wild is an important and formidable task.

Primary resources

Paper ↗Code ↗Data ↗

Benchmark Radar records only publicly supported details and links back to primary sources for verification.