task_three_d__object_scene__line_side_label

Summary

• Domain: three_d
• Scene id: object_scene
• Package: src/trace_tasks/tasks/three_d/object_scene/
• Supported query_id: left_of_directed_line, right_of_directed_line
• Answer type: option_letter
• Annotation type: point
• Annotation schema: point

Program Contract

Program: select(label(marked_points, side_of_directed_line(reference_object_a, reference_object_b) = requested_side)); scene=object_scene; scope=line_side_label

Candidate set: the visible 3D objects, surfaces, room/street/warehouse structures, spatial anchors, markers, and labeled options inside the line_side_label objective scope. Operands: visible scene state and prompt-bound operands named by label, marked_points, side_of_directed_line, reference_object_a, reference_object_b, requested_side, object_scene, line_side_label plus the active query_id branch. Operation: evaluate select over the candidate set using the finalized 3D scene state, camera projection, object identities, spatial relations, counts, distances, or option-selection constraints encoded in the program expression; generation enforces a unique final answer. Output binding: answer uses the option_letter schema; generation binds a unique final answer. Annotation witnesses: annotation uses the point schema; the prompt/annotation contract defines the minimal visual witnesses. Query ids: left_of_directed_line, right_of_directed_line.

Reasoning Operations

Families: spatial_relations

Contract

The image uses the object_scene renderer: a perspective 3D floor, table, or platform scene with one or more large context props, several uniquely named small objects, and six lettered point markers. Each marked point is rendered as a visible point glyph with a nearby letter label. The prompt names two unique reference objects and asks for the marked point on the requested side of the directed line from the first named object to the second.

The verifier computes the answer from finalized object projections and marker projections, not from pixels. The correct marker is the only marked point on the requested side of the projected directed reference line with a required side-distance margin. Distractor markers are constrained to the opposite side of that line. Render style, camera, canvas preset, object placement, labels, colors, and prompt wording variants are generation metadata, not public task axes.

Annotation Contract

Annotation is a scalar point at the selected marked point center. The selected point-glyph center is the only visual witness; marker label text and the reference object names are used only to identify the answer option and directed-line relation.

Prompt And Trace

The trace records selected prompt keys, camera/projection data, object and marker records, rendered pixel witnesses, answer-support metadata, reference-object roles, requested side, and solver fields needed to recompute the answer and annotation from the same finalized scene.

Determinism

Generation is deterministic from instance_seed, explicit params, config defaults, prompt bundle, and code versions. Answers and annotation come from the same finalized 3D scene trace.