Meitun LongCat Launches LoHoSearch as BrowseComp Scores Drop Below 30%
Search agent capabilities have been largely defined by BrowseComp over the last year. Yet, this benchmark is losing its relevance — it drove model performance from 30% to 90% in just ten months, and its value is rapidly diminishing. On July 17, Meituan’s LongCat launched a more rigorous evaluation, LoHoSearch, designed to reignite meaningful progress in search agent development.

LoHoSearch’s challenges are not manually crafted but automatically generated from a Wikipedia knowledge graph containing 7.62 million entities. Because they are machine-generated, these tasks reach levels of search space and structural complexity that human annotators could never consistently replicate. In short, while previous benchmarks were limited by human-designed ceilings, LoHoSearch delegates question generation to the knowledge graph, effectively removing that upper bound.

The results are stark. Among 11 leading models, the highest score reached only 34.74%, with the next three scoring between 15% and 16%. By contrast, these same models achieve approximately 90% on BrowseComp. This gap highlights the true limitations of current search agents. Notably, context strategies yield diminishing returns: the best context approach on LoHoSearch improved scores by just 6.8 percentage points, compared to 14 percentage points on BrowseComp. This suggests that relying on prompts and context engineering to bridge capability gaps is largely ineffective under this new benchmark.
LoHoSearch comprises 544 questions spanning 11 domains, structured in tree and graph formats, and has been open-sourced. As older benchmarks plateau, the real challenges for search agents are just beginning, and LoHoSearch may soon become an essential evaluation standard.
Related article
MiniMax Unveils H3 Multimodal Model, Cutting 15-Second 2K Audio-Visual Generation Costs to Under One-Third of Industry Standard
MiniMax has officially unveiled MiniMax H3, a general-purpose multimodal generation model designed to unify the understanding and creation of text, images, video, and audio. By bridging the historical divide between tasks and modalities in video gene
World Bank Urges Emerging Economies to Skip Big Models, Focus on Local Small Tools
The World Bank’s latest report charts a pragmatic course for developing economies: integrate artificial intelligence into public administration and commercial activities immediately, while tailoring existing solutions to local contexts. There is no n
OpenAI CEO Altman Cancels South Korea and Japan Visit Following the Birth of His Second Daughter
OpenAI CEO Sam Altman recently welcomed his second daughter, who was born prematurely, prompting the cancellation of his planned trip to South Korea and Japan scheduled for the 15th. While industry observers had speculated that the itinerary change m
Related Special Topic Recommendations
Comments (0)
0/500
Search agent capabilities have been largely defined by BrowseComp over the last year. Yet, this benchmark is losing its relevance — it drove model performance from 30% to 90% in just ten months, and its value is rapidly diminishing. On July 17, Meituan’s LongCat launched a more rigorous evaluation, LoHoSearch, designed to reignite meaningful progress in search agent development.

LoHoSearch’s challenges are not manually crafted but automatically generated from a Wikipedia knowledge graph containing 7.62 million entities. Because they are machine-generated, these tasks reach levels of search space and structural complexity that human annotators could never consistently replicate. In short, while previous benchmarks were limited by human-designed ceilings, LoHoSearch delegates question generation to the knowledge graph, effectively removing that upper bound.

The results are stark. Among 11 leading models, the highest score reached only 34.74%, with the next three scoring between 15% and 16%. By contrast, these same models achieve approximately 90% on BrowseComp. This gap highlights the true limitations of current search agents. Notably, context strategies yield diminishing returns: the best context approach on LoHoSearch improved scores by just 6.8 percentage points, compared to 14 percentage points on BrowseComp. This suggests that relying on prompts and context engineering to bridge capability gaps is largely ineffective under this new benchmark.
LoHoSearch comprises 544 questions spanning 11 domains, structured in tree and graph formats, and has been open-sourced. As older benchmarks plateau, the real challenges for search agents are just beginning, and LoHoSearch may soon become an essential evaluation standard.
MiniMax Unveils H3 Multimodal Model, Cutting 15-Second 2K Audio-Visual Generation Costs to Under One-Third of Industry Standard
MiniMax has officially unveiled MiniMax H3, a general-purpose multimodal generation model designed to unify the understanding and creation of text, images, video, and audio. By bridging the historical divide between tasks and modalities in video gene
World Bank Urges Emerging Economies to Skip Big Models, Focus on Local Small Tools
The World Bank’s latest report charts a pragmatic course for developing economies: integrate artificial intelligence into public administration and commercial activities immediately, while tailoring existing solutions to local contexts. There is no n
OpenAI CEO Altman Cancels South Korea and Japan Visit Following the Birth of His Second Daughter
OpenAI CEO Sam Altman recently welcomed his second daughter, who was born prematurely, prompting the cancellation of his planned trip to South Korea and Japan scheduled for the 15th. While industry observers had speculated that the itinerary change m





Home






