Skip to content

task_icons__named_field__counterfactual_total_count

Identity

  • domain: icons
  • scene_id: named_field
  • task: named_shape_counterfactual_count
  • module: src/trace_tasks/tasks/icons/named_field/counterfactual_total_count.py
  • prompt bundle: src/trace_tasks/resources/prompts/icons/named_field/icons_named_field_v1.json

Program Contract

Program: count.counterfactual(scene=named_field, scope=visible_icons, edit=shape_removal, target=total_icon_count, output=count)

Candidate set: the visible icon instances, icon attributes, fields, grids, paths, panels, reference items, and labeled option cards inside the visible_icons objective scope. Operands: visible scene state and prompt-bound operands named by named_field, visible_icons, edit, shape_removal, total_icon_count. Operation: evaluate count.counterfactual over the candidate set using the visible icon attributes, positions, relationships, transforms, counts, comparisons, 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 bbox_set schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: single.

Reasoning Operations

Families: counting, state_update

Scene And Query

The task renders one panel containing procedurally generated named shape icons. The prompt describes a hypothetical edit and asks for the integer count after applying that edit mentally.

Supported public query ids: - single

The fixed internal hypothetical branch is total_count_after_shape_removal. Trace metadata records it as query_spec.internal_query_id and query_spec.params.internal_query_id; it selects the shape-removal edit.

The target shape support is the full procedural named-icon vocabulary in src/trace_tasks/tasks/icons/shared/procedural_named_icons.py. Icons also sample a rendered fill_style from solid, striped, and dotted; fill style is metadata/visual variation only for this task and does not affect the hypothetical edit semantics.

Answer Contract

  • answer_gt.type = integer
  • default answer support is 1..8
  • value is the final count after applying the stated hypothetical edit

Annotation Contract

  • annotation_gt.type = bbox_set
  • one [x0, y0, x1, y1] pixel bounding box for every visible icon that contributes to the final count
  • annotation boxes are sorted by the witnesses in reading order
  • hypothetical additions are intentionally not sampled because they would not have visible bbox annotation

Trace Contract

  • scene_ir.entities contains one entity for each rendered procedural icon, including its fill_style.
  • scene_ir.relations.role_by_instance_id records each icon's counterfactual role and whether it is counted after the edit.
  • query_spec.params.query_id is single.
  • query_spec.params.internal_query_id records the fixed hypothetical branch.
  • execution_trace.target_answer equals the integer answer.
  • render_map.counted_instance_ids, witness_symbolic.counted_instance_ids, and projected_annotation.bbox_set are derived from the same rendered instances.

Prompt Contract

  • scene_key = single_scene_counting
  • task_key = counting_query
  • prompts ask for the resulting total icon count after a removal edit
  • Both output modes include contract-valid JSON examples