Skip to content

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

  1. Entities/relations: A visible Xiangqi-like line board with horizontal and vertical grid lines.
  2. Supported query_id values: horizontal_line_count, vertical_line_count
  3. answer_gt.type: integer
  4. Annotation schema: segment_set
  5. Alternate annotation forms: none
  6. Annotation witness policy: image-pixel line segments for all counted visible board lines; segment-set cardinality equals answer.
  7. Overlap/touch policy: each segment runs along the visible counted line from one board edge to the other.

3) Prompt contract

  1. prompt_bundle_id: games_counterfactual_board_v1
  2. scene_key: counterfactual_board
  3. task_key: board_grid_count_query
  4. Prompt query keys: horizontal_line_count, vertical_line_count
  5. Required slots: output-mode slots from the prompt bundle only.
  6. JSON example validity rule: segment-set cardinality equals the integer answer.
  7. Output modes: answer_only, answer_and_annotation

4) Determinism + constraints

  1. Seed namespaces used: task-local query/dimension namespaces plus shared counterfactual-board layout/style/noise namespaces.
  2. Unique-answer policy: the sampled visible horizontal or vertical line count directly determines the answer.
  3. Reject/resample conditions: invalid explicit style, dimensions, or query values raise and retry/propagate.
  4. No-auto-relaxation guarantee: semantic constraints are not relaxed.