HaiAI123

Curated Global AI Tools Directory

NewHot story
94
heat index
New

Meituan open-sources LoHoSearch, a search-agent benchmark

10/03/2026 — 10/03, 11:12·1 sources·1 reports

Story overview

On October 3, 2026, Meituan's technical team announced LoHoSearch, an evaluation benchmark for search agents, and released it as open source. The team frames it as a next-generation benchmark for search agent evaluation that uses a knowledge graph to calibrate how AI capability is understood.

In the same announcement, the team looked back at how Search Agent capability has evolved over the past year. On benchmarks such as BrowseComp, top models climbed from accuracy in the 30% range to above 90%. From that, the team concludes that once a benchmark saturates this quickly, its value for distinguishing between models declines — the context it offers for building LoHoSearch.

That is where the public information currently stops. The benchmark is open source, but the available material does not describe its structure, how many tasks it contains, or which dimensions it measures, and it lists no model scores on the new benchmark. The knowledge-graph element appears only in that single phrase, with no further explanation of how it is used.

AI-generated from 1 reports · updated 2 hours ago

Latest turnMeituan has open-sourced LoHoSearch, a next-generation benchmark for search agents that uses a knowledge graph to calibrate how AI capability is measured. The team notes that on benchmarks such as BrowseComp, top model accuracy climbed from the 30% range to over 90% in the past year, eroding their value for telling models apart.

Reports on this story headlines open the original

Today
  1. Meituan has open-sourced LoHoSearch, a next-generation benchmark for search agents that uses a knowledge graph to calibrate how AI capability is measured. The team notes that on benchmarks such as BrowseComp, top model accuracy climbed from the 30% range to over 90% in the past year, eroding their value for telling models apart.

    美团技术团队AI score 74

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to the hot board →