SuperGlasses : Benchmarking Vision Language Models as Intelligent Agents for AI Smart Glasses

Zhuohang Jiang, Xu Yuan, Haohao Qu, Shanru Lin, Kanglong Liu, Wenqi Fan, Qing Li, [0.1cm] The Hong Kong Polytechnic University City University of Hong Kong · 2026

Read SuperGlasses : Benchmarking Vision Language Models as Intelligent Agents for AI Smart Glasses by Zhuohang Jiang, Xu Yuan, Haohao Qu, Shanru Lin, Kanglong Liu, Wenqi Fan, Qing Li, [0.1cm] The Hong Kong Polytechnic University City University of Hong Kong (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-22683.txt.

Description

Zhuohang Jiang and six co-authors benchmark vision-language models as smart-glasses assistants. AI glasses are a hot wearable, and answering questions about the world from what the wearer sees is a core use. Yet standard multimodal datasets miss the glasses setting, where finding the object of interest must come before looking up outside knowledge.

SUPERGLASSES, built purely from real glasses-collected data, holds over 2,400 egocentric image-question pairs across 14 image domains and 8 query types, with search paths and reasoning notes. Testing 26 leading models exposed wide gaps. The team's own agent, SUPERLENS, couples object detection, query splitting, and multimodal web search, beating GPT-4o by about two percent and setting the benchmark's top score.

Catalog record

CREATORZhuohang Jiang, Xu Yuan, Haohao Qu, Shanru Lin, Kanglong Liu, Wenqi Fan, Qing Li, [0.1cm] The Hong Kong Polytechnic University City University of Hong Kong
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE70 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-22683

Browse the full library · About the Tower of Babel