ContrAR: Dataset for Contradictory Virtual Content Attacks in Augmented Reality

This is the official dataset of the paper: Benchmarking Vision-Language Models under Contradictory Virtual Content Attacks in Augmented Reality. The paper was accepted to CVPR 2026 Findings.

This dataset contains augmented reality videos collected to evaluate the detection of contradictory virtual content in AR. It covers five application scenarios: Indoor Navigation, Outdoor Navigation, Safety Inspection, Smart Apartment, and Smart Retail.

Each video is labeled as either "contradictory" (with the video file name starts with "A") or "not contradictory" (with the video file name starts with "N"), indicating whether different pieces of virtual content present semantically conflicting information.

More details about the dataset can be found at: https://github.com/YM-Xiu/ContrAR-Dataset