aifollow.news
Back Rohan Paul
Rohan Paul· @rohanpaul_ai · X· · Original publication time AI score54

Harvard study finds no clear drop in code merge rates during Claude Fable 5 suspensionMachine translation

AI introduction

A Harvard study examined a 19-day suspension of Claude Fable 5 under a U.S. export control directive, comparing 297 former users with 8,908 developers who remained on Opus 4.8. Access to Fable 5 raised the seven-day pull-request merge rate by just 1.4 percentage points, a change the study describes as within noise; 94.5% of pull requests had already been merged beforehand.

Original post · AI translation

@rohanpaul_ai
Translation · refer to the original

哈佛大学的一篇新论文发现,在Claude Fable 5被停用的19天里,其用户改用性能较弱的Opus 4.8后,拉取请求获得合并的频率与此前大致相同。

2026年6月,美国一项出口管制指令使Claude Fable 5中断服务19天,大多数用户转而使用Opus 4.8。该研究追踪了297名Fable 5用户的GitHub拉取请求,并将其与持续使用Opus 4.8的8,908名开发者进行了比较。

使用Fable 5使7天内的合并率仅提高了1.4个百分点,这一变化处于随机波动范围内。在另外5次模型升级中,合并率也没有上升,不过此前已有94.5%的拉取请求获得合并。

Source: Rohan Paul · x.com

Research
Found an error? Send a correction