Promptable Gaze Target Estimation via Natural Language
August 10, 2026
A new Promptable Gaze Target Estimation (PGE) paradigm enables end-to-end gaze prediction conditioned on natural language prompts. This eliminates the need for brittle, multi-stage pipelines requiring explicit head bounding boxes or pose inputs.
HOW THIS AFFECTS YOU
●
builderYou can now specify gaze analysis tasks using text prompts rather than complex pre-processing pipelines.
●
researcherThis introduces a new concept-driven paradigm for evaluating gaze analysis flexibility.