task_illustrations__construction_site__worker_attribute_count

Summary

• Domain: illustrations
• Scene id: construction_site
• Implementation scene: construction_site
• Implementation source: src/trace_tasks/tasks/illustrations/construction_site/worker_attribute_count.py

Contract

1. Domain: illustrations
2. Scene id: construction_site
3. Public task id: task_illustrations__construction_site__worker_attribute_count
4. Supported query_id values: hard_hat_color_worker_count, vest_color_worker_count
5. Query ids: hard_hat_color_worker_count, vest_color_worker_count
6. Answer schema: integer_count
7. Annotation schema: bbox_set
8. Program schema: count(filter(workers, worker_selector(worker, target_attribute, target_attribute_value))); scene=construction_site; scope=worker_attribute_count

Program Contract

Program: count(filter(workers, worker_selector(worker, target_attribute, target_attribute_value))); scene=construction_site; scope=worker_attribute_count

Candidate set: the visible illustrated scene objects, people, regions, tiles, patches, labels, and option panels inside the worker_attribute_count objective scope. Operands: visible scene state and prompt-bound operands named by filter, workers, worker_selector, worker, target_attribute, target_attribute_value, construction_site, worker_attribute_count plus the active query_id branch. Operation: evaluate count over the candidate set using the visible illustrated objects, regions, layout relationships, counts, patch/tile transforms, or option-selection constraints encoded in the program expression; generation enforces a unique final answer. Output binding: answer uses the integer schema; a non-negative integer derived from the same execution trace as the annotation. Annotation witnesses: annotation uses the bbox_set schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: hard_hat_color_worker_count, vest_color_worker_count.

Reasoning Operations

Families: filtering, counting

Task Contract

Counts visible workers matching one sampled safety-gear color attribute.

Query Branches

Query id	Program schema
hard_hat_color_worker_count	count(filter(workers, worker_selector(worker, target_attribute, target_attribute_value))); scene=construction_site; scope=worker_attribute_count; query_branch=hard_hat_color_worker_count
vest_color_worker_count	count(filter(workers, worker_selector(worker, target_attribute, target_attribute_value))); scene=construction_site; scope=worker_attribute_count; query_branch=vest_color_worker_count

Program Metadata

• Program signatures: count.single_attribute_membership
• Base program contract: count(filter(workers, worker_selector(worker, target_attribute, target_attribute_value))); scene=construction_site; scope=worker_attribute_count
• Parameter axes: target_attribute
• Arguments:
• target_attribute: object_attribute; allowed hard_hat_color, vest_color; source program_schema_concrete|query_id|parameter_axes
• target_attribute_value: object_attribute; allowed sampled_color; source program_schema_concrete
• worker: semantic_role; allowed worker_instance; source program_schema_concrete
• workers: semantic_role; allowed visible_workers; source program_schema_concrete
• Argument metadata status: curated
• Supported query ids: hard_hat_color_worker_count, vest_color_worker_count

Answer Contract

• Answer schema: integer_count
• Generator answer_gt.type: integer
• Configured answer support is 0..5.
• The answer value is a non-negative integer derived from the same execution trace as the annotation.

Annotation Contract

• Annotation schema: bbox_set
• Generator annotation_gt.type: bbox_set
• Annotation is an unordered set of final-image pixel points, one near the center of each counted worker; use an empty set when the answer is 0. Do not include labels, numeric annotations, or context-only regions.
• Annotation and answer must be projected from the same generated scene trace, not inferred from pixels or prompt text.

Prompt And Trace Requirements

• Prompt text must come from the illustrations_construction_site_v1 scene prompt bundle, with scene/task/output layers selected deterministically and recorded in metadata.
• Render randomness, sampled fonts/styles, query operands, and verifier payloads must be explicit in the instance trace.
• Distractor/context text may be rendered only when it is part of the scene grammar and must not be treated as annotation unless it is the queried visual witness.