UIST '26 · Detroit

ContextXR

A benchmarking framework for context-aware XR interfaces, built on functional facets: groups of capabilities that together serve one user intent.

Hyunsung Cho, Sarah Yewon Yun, Nancy Ruonan Sun, Ben Lafreniere, Mark Parent, Kashyap Todi, Tanya R. Jonker, Hrvoje Benko, Tongshuang Wu, David Lindlbauer

Carnegie Mellon University · Reality Labs Research, Meta

MineXR++ Facet Graphs

Each app's functionality is a graph of facets. Edges are open_<facet> capabilities that move from one facet to another. Click an app to expand its graph, and click a facet to see its capabilities. Hover a facet to preview its panel, and open the panel to see where each capability sits.

MineXR++ Contextual Layouts

Each MineXR scenario is one participant's XR layout for one task in one environment. Pick an environment, then one of its layouts, grouped by the kind of task. Drag a widget to rearrange it, and click it to see which of its capabilities the participant's widgets showed. Below the view, test a functionality suggestion method on the same layout.

Context Factor Analysis

Which part of the context predicts a facet: who the user is, what they are doing, or where they are? For each facet, lift compares how often it appears in both layouts of a pair that share the same user (or task, or environment) with how often it appears in both layouts of any pair. Above 1, that factor makes the facet more likely. A facet's dominant factor has the highest lift; it is mixed when the top two are within 1.2× of each other, and has no lift when none is above 1.

· click a group to filter

Each dot is a facet, pulled toward the factors with more lift and sized by how many layouts it appears in. Dots are nudged apart so none hides another. Labelled: the paper's examples (Fig. 4).

Lift by factor, scaled to each facet's highest · tick: lift 1 (no effect) · number: placed in N layouts