PInVerify: An Offline Embodied Benchmark for Active Instance Verification

Yuhang Jiang · 2026

Read PInVerify: An Offline Embodied Benchmark for Active Instance Verification by Yuhang Jiang (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-2605-30639.txt.

Description

Jiang tackles a gap in embodied AI: reaching a target's area does not prove the agent found the right instance, since details like floral versus striped patterns demand close inspection. He formalizes Active Instance Verification: an agent picks viewpoints around a candidate to judge whether it matches a detailed description. His PInVerify benchmark offers thousands of episodes across many categories, with captures, six sectors, trap views, and unreachable sectors. The best multimodal baseline beats the embedding baseline, active viewpoint choice shows no gain, and a fine-tuned agent scores high.

Catalog record

CREATORYuhang Jiang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE111159
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.30639

Browse the full library · About the Tower of Babel