Papers & Research

Meituan LongCat releases LoHoSearch: a harder search agent benchmark

Original title:BrowseComp went from 30% → 90% in 10 months. Search-agent benchmarks are saturating. So we built Lo...

AI summary

Meituan LongCat has launched LoHoSearch, a search agent benchmark that automatically generates questions from a Wikipedia knowledge graph of 7.62 million entities, aiming to address the saturation of existing benchmarks such as BrowseComp. In tests on 11 frontier models, the best score was only 34.74%, far below the roughly 90% current models achieve on BrowseComp; context strategies brought only a +6.8 percentage point improvement. The benchmark includes 544 questions across 11 domains, uses tree and graph structures, and has been open-sourced.

Key Takeaways

  • Meituan LongCat has launched LoHoSearch,a search agent benchmark that automatically generates…
  • 62 million entities,aiming to address the saturation of existing benchmarks such as BrowseComp.
  • In tests on 11 frontier models,the best score was only 34.
  • 74%,far below the roughly 90% current models achieve on BrowseComp
  • context strategies brought only a +6.

Quick Facts

Published
Updated
Source
X:美团 LongCat (@Meituan_LongCat)
Category
Papers & Research
AI score
75
Read the original Read the full article at the original source
Why it matters

BrowseComp 快被刷到顶了,LongCat 甩出 LoHoSearch,前沿模型集体回落到三成出头——搜索 agent 的真正挑战才刚开始,这个基准可能是下一个必测项。

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: Meituan LongCat releases LoHoSearch: a harder search agent benchmark. July 17, 2026. https://new.nat6.net/en/news/cmrp0qw3s090ybitorwj6diwi/

View all →