Original title: GazeFS: Target-Centered Gaze-Trajectory Forecasting and Stabilization from Gaze-Head History
Yaozheng Xia, Zaiping Zhu, Bo Pang, Minghao Xie +3 · September 2026
Using your eyes to select targets on a screen is tricky. Real gaze jitter isn't just random noise; reaching a target alters how your eyes and head move, often leaving a steady aiming error. Researchers created GazeFS to fix this. It watches a moving history of past eye and head movements. From those patterns, it predicts the true target-center direction and guesses whether you are searching or focusing, all without knowing where the actual targets are located. The system cuts down pointing bias and reduces target error during focus periods. Looking at past movement history decoded the user's current phase better than just checking where the eyes pointed at that single instant, showing that past head and eye coordination reveals where you truly intend to look.
Imagine you are throwing a ball to a friend who keeps shifting slightly to one side. If you aim only at where they are now, you’ll miss often because they move between when you see them and when the ball arrives. Here, the system learns from your past throws and body movements to predict where you should aim next, even without seeing your friend’s current position, by recognizing patterns in how they move relative to you.
This matters for people using eye-tracking in devices like virtual reality or assistive technologies. It makes selecting on-screen targets more accurate and stable, reducing errors that can frustrate users or hinder performance.
The abstract doesn't establish limitations explicitly, but it focuses on specific acquisition episodes and shows improvements in targeted metrics. It notes that temporal smoothness is a separate objective not addressed by this work. Real-world generalizability beyond the tested participants and conditions is not confirmed.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
GazeFS: Target-Centered Gaze-Trajectory Forecasting and Stabilization from Gaze-Head History — Yaozheng Xia, Zaiping Zhu, Bo Pang et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.