新闻
VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use Policies
arXiv cs.AI · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- VAKRA benchmark evaluates AI agents on multi-hop reasoning across 8,000+ executable APIs in 62 domains with tool-use policy constraints.
- 为何重要
- Enterprise engineers building or deploying AI agents that must combine API calls and document retrieval under operational constraints.
- 注意
- Best models achieve only 70% on single-hop tasks and 50% on multi-step reasoning; failures stem from language reasoning, not tool mechanics.
收听本摘要
- agent
- reasoning
- retrieval
- tool-use
- edge
这条新闻背后的模式
每个模式都讲清楚技术如何运作、何时值得投入，以及在哪里会失效。
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。