PapersSelected

NLAH论文:用自然语言替代智能体代码,Token减少95%

Why we picked it

看这篇论文,阿里云用自然语言写智能体工具,token省了95%,还能揪出拖后腿的环节。

阿里云等机构在论文中提出NLAH框架,用可执行自然语言替换刚性代码工具集(Agent Harness)。在相同任务上,NLAH性能与代码方案持平,但Token消耗从60k降至2.

Body

9k,降幅达95%。其模块化设计可精准归因每一步的值,并识别出多候选搜索等“负资产”环节。

该工作将智能体构建从“胶水代码”转向科学策略。

Original · 阿里云 Alibaba Cloud

🚀 Taming Agent Chaos?

Paper reveals NLAH: Replace rigid code harnesses with executable natural language.

✅ Performance matches code, tokens drop 95% (60k→2.9k) ✅ Modular design enables precise value attribution ✅ Identifies "negative assets" like multi-candidate search

Shift from glue code to scientific strategy.

💡https://t.co/0YRLpGUF9m

#AgentHarness #NLAH #LLMEngineering

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk. Key entities and technical topics under continuous tracking include NLAH, AgentHarness, 智能体, 自然语言.

Source attribution: curated from 阿里云 Alibaba Cloud (@alibaba_cloud).