task_symbolic__logic_gate_circuit__internal_output_count

1) Identity

1. Domain: symbolic
2. Scene id: logic_gate_circuit
3. Task id: task_symbolic__logic_gate_circuit__internal_output_count
4. Objective contract: internal_output_count

Program Contract

Program: logic_gate_circuit.internal_output_count(scene=logic_gate_circuit, scope=single_circuit_with_input_values, target_output=0|1, output=integer)

Candidate set: all visible gate symbols in the single rendered circuit. Operands: the visible input values, gate topology, each gate's operator, and the requested local output value. Operation: evaluate every gate's local output and count gates whose output equals the requested value. Output binding: answer is the matching gate count as an integer. Annotation witnesses: a homogeneous bbox_set of gate-symbol bboxes whose computed output matches the requested value. Query ids: internal_output_one_count, internal_output_zero_count.

Reasoning Operations

Families: filtering, counting, logical_composition, topology, formula_evaluation

2) Scene + task contract

1. Entities/relations: one fanout-free expression-tree circuit with visible input values, standard gate symbols, wires, and a final OUT node.
2. Supported query_id values: internal_output_one_count, internal_output_zero_count
3. answer_gt.type: integer
4. Default annotation_gt.type: bbox_set
5. Annotation schema: bbox_set
6. Annotation witness policy: annotation marks the gate-symbol bounding boxes whose computed output equals the requested value; nonmatching gates, input labels, wires, output node, panel borders, and decorative background are not prompt-facing annotation.
7. Overlap/touch policy: gate-symbol bounding boxes are laid out by the shared expression-tree renderer and should not overlap.

Query id	User-facing operation
internal_output_one_count	Count gate symbols whose computed local output is 1.
internal_output_zero_count	Count gate symbols whose computed local output is 0.

3) Prompt contract

1. prompt_bundle_id: symbolic_logic_gate_circuit_v1
2. scene_key: logic_gate_circuit
3. task_key: logic_gate_internal_output_count
4. Query keys: internal_output_one_count, internal_output_zero_count
5. Required slots:
6. answer output: json_output_contract_answer_only, answer_hint, json_example_answer_only
7. answer+annotation mode: json_output_contract, annotation_hint, answer_hint, json_example
8. JSON examples use an integer answer and a bbox_set annotation array.
9. Output modes: answer_only, answer_and_annotation

4) Determinism + constraints

1. Seed namespaces used: <task_id>.dataset and <task_id>:scene_variant.
2. Unique-answer policy: the sampler chooses a target answer and resamples four-gate circuits until the per-gate evaluation trace has exactly that many gates with the requested output.
3. Reject/resample conditions: invalid query id, target answer outside 0..gate_count, or incompatible gate-count settings.
4. No-auto-relaxation guarantee: semantic constraints are validated and never relaxed after sampling.

5) Tests

1. Determinism test: covered by source-layout generation tests.
2. Answer/annotation consistency test: tests/test_symbolic_logic_gate_tasks.py
3. Prompt metadata/placeholder test: tests/test_prompt_system.py
4. Constraint-specific tests: focused logic-gate task tests assert per-gate output counts and bbox_set annotation cardinality.