task_games__rule_override_board__line_result_count¶
Program Contract¶
Program: count(filter(mini_boards, anti_line_result(target_player)=target_result)); scene=rule_override_board; scope=line_result_count
Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the line_result_count objective scope.
Operands: visible scene state and prompt-bound target operands named by the task contract.
Operation: evaluate count(filter(mini_boards, anti_line_result(target_player)=target_result)) over the visible mini-board set; generation enforces a unique final answer.
Output binding: answer uses the integer schema; generation binds a unique final answer.
Annotation witnesses: annotation uses the bbox_set schema for every counted mini-board panel.
Query ids: line_override_win_count, line_override_loss_count.
Reasoning Operations¶
Families: filtering, counting
Generation Notes¶
- The prompt states the anti-line rule: a full row, column, or diagonal is a loss for the target player.
line_override_win_countcounts mini-boards where the target player wins;line_override_loss_countcounts losses.- Annotation bboxes are the counted mini-board panels.
- The answer and annotation are bound from the same generated board state.