On BFCL v3 test sample 051, when asked for an airfare quote, the unguided baseline improperly proceeds into booking, authentication, and card operations, causing a state mismatch error. In contrast, the PG-guided agent correctly finishes the turn upon providing the quote without executing unrequested follow-up booking actions. · CiteArk
On BFCL v3 test sample 051, when asked for an airfare quote, the unguided baseline improperly proceeds into booking, authentication, and card operations, causing a state mismatch error. In contrast, the PG-guided agent correctly finishes the turn upon providing the quote without executing unrequested follow-up booking actions.
