mixed reality developers

Abstract: Mixed reality interfaces must continuously adapt to a user's changing task and environment, yet adaptation policies are fixed in advance, either preset by the system or configured by the user. Natural language offers a way to state changing needs mid-task, but whether users can express interface behavior this way, and what it gains over direct manipulation, is unknown. We first elicited how users verbalize adaptation needs (N=19). We then built ConverseMR, which translates a spoken request into adaptation objectives and constraints and satisfies them continuously as the user's context changes. In a within-subjects study (N=24), conversational and direct control proved complementary: direct manipulation for precision, ConverseMR for reach and expressive breadth, with their combination rated highest. Spoken commands were rarely final and users frequently refined and replaced them upon experiencing the result, characteristic of, what we call, user-led adaptation, where users communicate adaption needs and the system identifies a solution.

Abstract: We propose SnapPhysics, a training-free framework that reconstructs 3D objects and estimates their physical properties such as mass, friction, and center of gravity from a single image. For physically coherent interactions in mixed reality (MR), such properties are as important as geometry. Prior approaches infer them by analyzing object dynamics in video, which is computationally costly, or by querying vision-language models (VLMs) on single images, which lacks geometric grounding and inter-object relationships. We address these limitations by combining instance-level 3D reconstruction and spatial alignment with a physics-aware scene graph that encodes these relationships and per-object metric geometry as structured context for VLM-based property reasoning. Experiments on 3D-FRONT show that SnapPhysics improves scene-level F-Score by 18.6% over the best learning-based method, and on real captured scenes with ground-truth mass, it reduces the mean absolute log difference error (mALDE) by up to 20.5% and improves log-scale correlation ($r^2_{\mathrm{ls}}$) by up to 19.6% over VLM-only estimation. SnapPhysics enables physically interactive MR experiences without manual parameter tuning. Project page: https://snapphysics-ismar2026.github.io/.