Live page · Day archive

HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents

Read the original at HF Daily Papers

Summary

Researchers introduce HyperBrowseComp, a multilingual and multimodal benchmark containing 423 manually authored questions that test web-browsing agents on locating obscure evidence across 13 languages.

Carried by: HF Daily Papers. First seen: .