capability

Overview ¶

Package capability is Pluto's core-capability qualification pack: instruction-following and known-answer cases that need no special target capability, just plain text conversation. All evaluators are programmatic.

Each case is its own single-scenario table: exact.RequiredText/ForbiddenText are configured per evaluator (not per scenario), and a table shares one evaluator set, so per-case expectations map onto one table per case rather than a dispatching meta-evaluator.