Quoting Prompting Claude Fable 5.2
Quoting Prompting Claude Fable 5.2
Anthropic's documentation notes that Claude Fable 5.2 is highly sensitive to how a prompt describes code ownership. In tests on small coding tasks, when a prompt said nothing about ownership or presented the codebase as someone else's, the model tended to follow the existing structure, even when that repeated existing code. When told explicitly that the codebase is its own, the model merged repeated code in about half of tries, against under a tenth without that prompt.
Selection record
Not admittedSum of both 80 < twice the threshold 152
- Source tier
- Media and individuals; this tier's threshold is 76
- Pre-filter
- passed：引用Claude模型提示词与代码归属测试
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Simon Willison · simonwillison.net