Concept智能体基准测试
Agent benchmark
A standardized test suite for evaluating the capabilities of AI agents, typically consisting of predefined tasks, scoring rules and reference results. Its core metric is task success rate, but it seldom records the behavioral trajectories of the process.
Timeline (last 90 days)
Oct 27
The core metric of traditional agent benchmarks is task success rate, and they rarely record what the agent said or did to intermediaries, tools, or collaborators during execution.
Unverified50%