AITOPAI热报
ModelsOfficial

Nace AI 发布 Drex for Search：BrowseComp-Plus 得分 92.05%

Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls

官方一手marktechpost@Asif Razzaq
Blog/Media· 2 sources
Read the original
Why we picked it

Nace AI 把 Drex 决策模型用到智能体搜索上了，BrowseComp-Plus 跑到 92.05%，而且平均只用 12.6 次搜索调用，做搜索智能体的可以看看。

Nace AI 发布 Drex for Search，将其 Drex 决策模型引入 AI 智能体的文档搜索环节。官方报告在 BrowseComp-Plus 基准上取得 92.05% 的成绩，平均每次任务只需 12.6 次搜索调用。在 OfficeQA Pro 基准上，该模型还减少了智能体的对话轮数，降低了多步搜索的开销。

Image · marktechpost
Original · marktechpost

Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls

Nace AI has released Drex for Search, bringing its Drex decision models to document search for AI agents. Nace reports 92.05% on BrowseComp-Plus with 12.6 search calls, and fewer agent turns on OfficeQA Pro. The post Nace AI Releases Drex for Search for AI Agents: 92.05% on BrowseComp-Plus With 12.6 Search Calls appeared first on MarkTechPost .

#Drex for Search#Nace AI#BrowseComp-Plus#智能体

Keep reading

Recent stories with the same tag or section
ProductsBuildersSelected23:13
Nace AI launches Drex for Search for agent document retrieval

The gains come from cutting agent steps and input tokens by more than half, which lowers API costs on retrieval-heavy workloads rather than just raising accuracy a few points. Teams building agents over enterprise document stores can plug it into Claude Code, Codex or Hermes Agent as a tool. The OfficeQA benchmark result offers a direct comparison point for evaluating retrieval layers before committing to one.

#Drex for Search#Nace AI#Claude Code#Codex
TestingCatalog@testingcatalogSource
ModelsBuildersSelected00:01
Kaggle launches FrontierCS 2 benchmark with 48 research tasks

The setup measures genuine research-style problem solving rather than memorised answers, since the agents never see the paper behind the code they must beat. Researchers and agent builders can use the leaderboard to compare frontier models on open-ended engineering work, a dimension most coding benchmarks do not cover. Tasks remain open on Kaggle for anyone to attempt.

#Kaggle#FrontierCS#benchmark#agents
Kaggle@kaggleSource