task_games__counterfactual_board__board_line_count¶
Program Contract¶
Program: count(board_grid.visible_lines, orientation=horizontal|vertical, style=xiangqi); scene=counterfactual_board; scope=board_line_count
Candidate set: the visible game board, pieces, tokens, cards, tiles, marked state, legal-move cues, result panels, and labeled options inside the board_line_count objective scope.
Operands: visible scene state and prompt-bound operands named by board_grid, visible_lines, orientation, horizontal, vertical, style, xiangqi, counterfactual_board, board_line_count plus the active query_id branch.
Operation: evaluate count over the candidate set using the visible game state, rules, legal moves, comparisons, counts, simulations, or option-selection constraints encoded in the program expression; generation enforces a unique final answer.
Output binding: answer uses the integer schema; generation binds a unique final answer.
Annotation witnesses: annotation uses the segment_set schema; image-pixel line segments for all counted visible board lines; segment-set cardinality equals answer.
Query ids: horizontal_line_count, vertical_line_count.
Reasoning Operations¶
Families: counting
2) Scene + task contract¶
- Entities/relations: A visible Xiangqi-like line board with horizontal and vertical grid lines.
- Supported
query_idvalues:horizontal_line_count,vertical_line_count answer_gt.type:integer- Annotation schema:
segment_set - Alternate annotation forms: none
- Annotation witness policy: image-pixel line segments for all counted visible board lines; segment-set cardinality equals answer.
- Overlap/touch policy: each segment runs along the visible counted line from one board edge to the other.
3) Prompt contract¶
prompt_bundle_id:games_counterfactual_board_v1scene_key:counterfactual_boardtask_key:board_grid_count_query- Prompt query keys:
horizontal_line_count,vertical_line_count - Required slots: output-mode slots from the prompt bundle only.
- JSON example validity rule: segment-set cardinality equals the integer answer.
- Output modes:
answer_only,answer_and_annotation
4) Determinism + constraints¶
- Seed namespaces used: task-local query/dimension namespaces plus shared counterfactual-board layout/style/noise namespaces.
- Unique-answer policy: the sampled visible horizontal or vertical line count directly determines the answer.
- Reject/resample conditions: invalid explicit style, dimensions, or query values raise and retry/propagate.
- No-auto-relaxation guarantee: semantic constraints are not relaxed.