task_puzzles__word_search__search_location_label

Program Contract

Program: select_label(location_option, rule=target_word_start_cell_and_direction); scene=word_search; scope=search_location_label

Candidate set: the visible letter grid, target or option words, candidate location/direction labels, and highlighted word path inside the search_location_label objective scope. Operands: visible scene state and prompt-bound operands named by location_option, target_word_start_cell_and_direction, word_search, search_location_label. Operation: evaluate select_label over the candidate set using the visible states, constraints, transforms, comparisons, counts, paths, or option-selection rules encoded in the program expression; generation enforces a unique final answer. Output binding: answer uses the option_letter schema; the correct option label. Annotation witnesses: annotation uses the bbox_sequence schema; the ordered sequence of grid-cell bounding boxes for the found. Query ids: single.

Reasoning Operations

Families: spatial_relations, topology, matching

Answer And Annotation

1. answer_gt.type = option_letter
2. answer_gt.value is the correct option label.
3. annotation_gt.type = bbox_sequence
4. Annotation schema: bbox_sequence
5. Annotation is the ordered sequence of grid-cell bounding boxes for the found word, from first letter to last letter.
6. scalar_annotation_checked = true; this task has multiple ordered cell witnesses.

Query Contract

1. Public query_id: single
2. Option labels are visible answer candidates, not public query branches.
3. Internal variation covers word, direction, grid size, and scene styling only.

Prompt Contract

1. Bundle: puzzles_word_search_v1
2. Scene key: word_search
3. Task key: search_location_label_query
4. Prompt query key: search_location_label