Unverified50% confidenceFactExact time
The overall agent efficiency score proposed by Bland consists of a funnel model composed of three progressively layered ratios: autonomous completion rate, token efficiency ratio, and value generation ratio.
1
Sources
50%
Confidence
Long-term
Relevance
10/7/2026
First Seen
Sources
Related Entities
Related Claims
Unverified评估模型不能只看智能指标，还要看Token使用效率和工具调用次数，工具调用次数过多会导致上下文污染67% similarUnverified综合智能指数由智能体工作、编程、科学推理、通用能力等 9 项独立评估构成，Sol 与 Faber-5 几乎持平仅差 1 分，但任务完成时间减少约 61%，成本约为原来的一半67% similarUnverifiedThe algorithm employs variance-reduced estimation of the Bellman optimality operator locally at each agent.67% similarUnverified一些团队实际度量数据显示，AI带来的净效率提升在20%-40%之间，远低于营销材料声称的数倍提升66% similarUnverified2023年MIT与Google的联合研究表明，当子任务存在明确完成标准时，智能体的自我纠错效率可提升40%以上66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/984912API
curl https://kongchang.com/api/v1/knowledge/claims/984912MCP
get_claim(id=984912)