Tag
The article advocates for using a strict pass metric instead of partial credit to evaluate finance AI agents, referencing the Ling-3.0-flash-Fin release to improve reliability in complex, step-dependent tasks.