The company said that in testing, 95 percent of Fable sessions ran entirely on Fable responses, without falling back to Opus 4.8.
这个95%的统计数据需要进一步验证。测试样本大小、测试场景的代表性以及如何定义'完全运行'都值得深入了解。这个数据可能影响用户对模型可靠性的判断。
4 Matching Annotations
- Last 7 days
-
www.theverge.com www.theverge.com
-
- Apr 2026
-
www.anthropic.com www.anthropic.com
-
In our internal evals and testing, medium effort achieved slightly lower intelligence with significantly less latency for the majority of tasks.
大多数人认为内部评估和测试足以代表用户真实体验，但作者承认他们的内部测试未能准确捕捉到用户对AI智能度的实际感知差异。这暗示了实验室环境与实际使用场景之间存在根本性脱节，挑战了传统产品测试方法论的有效性。
-
- Oct 2020
-
twitter.com twitter.com
-
Health Nerd on Twitter. (n.d.). Twitter. Retrieved October 17, 2020, from https://twitter.com/GidMK/status/1316511734115385344
-
- Jul 2020
-
news.sky.com news.sky.com
-
Coronavirus: The inside story of how UK’s “chaotic” testing regime “broke all the rules.” (n.d.). Sky News. Retrieved July 17, 2020, from https://news.sky.com/story/coronavirus-the-inside-story-of-how-uks-chaotic-testing-regime-broke-all-the-rules-12022566
-