DP-LENS: VR lens with LLM auto-routing slashes cognitive load in 3D data

Exploring complex 3D data in virtual reality is powerful but problematic: dense datasets are heavily occluded, and traditional navigation tools require constant manual manipulation, driving up cognitive load and physical fatigue. DP-LENS, a new system from researchers Nieyu Cao, Xian Wang, and Lik-Hang Lee (accepted to ISMAR 2026), tackles this head-on. It combines a density-aware polyfocal fisheye lens—which deforms geometry to peek around obstacles while preserving peripheral context—with topology-driven auto-routing that intelligently charts a path through the data. A Large Language Model (LLM) acts as a supplementary voice-based target selector, letting users say “find the cluster of neurons” and letting the system handle the navigation automatically.

In the first user study (N=18), manual DP-LENS outperformed two industry-standard baselines—World-in-Miniature and volumetric slicing—on heavily occluded datasets. The lens significantly reduced cognitive load, decreased completion time, and earned higher user preference. The second study (N=16) added the LLM-initiated auto-routing system and compared it against fully manual DP-LENS. Auto-routing further improved task efficiency, lowered cognitive load, and boosted preference. Notably, it partially decoupled exploration efficiency from the physical dimensions of the data, meaning the size of the dataset mattered less, and it mitigated physical fatigue. The authors propose design implications for future scalable, low-fatigue 3D visual analytics systems, suggesting a shift toward spatially intelligent interfaces that combine AI reasoning with human spatial intuition.