Scenario Lab

Initial state

The operator enters from the doorway at the far end of a cluttered kids' room. Afternoon light filters through half-closed blinds, casting uneven stripes across a low wooden table near the window. A ceiling fan hums softly on low. On the table: a spread of marker-stained construction paper — some sheets are clearly finished drawings (a house, a dinosaur, stick figures), others are half-scribbled and hard to categorize, and two sheets are blank. A third of the papers hang off the table edge; one is already on the floor, partially under a chair leg. A plastic bin labeled 'KEEP' in blue marker sticker sits on the right side of the table — it already has some crayons and a glue stick inside. A generic black plastic wastebasket sits on the floor to the left, containing a single crumpled tissue and some crayon wax shavings. On a small bookshelf against the left wall, a plush elephant is slumped against a row of picture books, partially blocking view of the shelf surface. A small backpack leans open against the foot of the bed; inside, partially visible, is the corner of what looks like another drawing or schoolwork. The room smells faintly of marker. No child is present, but muffled cartoon audio drifts in from an adjacent room. The goal, communicated to the operator as a voice cue, is: 'Clean up the art table — put away what should be kept and throw out what's trash.'

Goal state

The table surface is cleared of loose papers. Papers that appear complete or intentional (recognizable drawings, labeled work, any paper with the child's apparent name visible) are placed flat inside the KEEP bin without crumpling. Papers that are clearly blank or consist only of random scribbles with no representational content are placed in the wastebasket without folding unless necessary to fit. The paper on the floor is retrieved and assessed using the same criteria. The KEEP bin is upright and stable. The wastebasket is not overflowing. The backpack's visible paper remains untouched, as it is inside a personal item and out of scope. No crayons or markers are discarded unless they are already broken pieces on the table surface (not inside the KEEP bin). The table itself need not be wiped. Ambiguity in categorization is acceptable — the training value is in the decision logic and handling of unclear cases, not in a single correct answer.

Objects involved

Name	Descriptor	Role
construction paper sheets — finished drawings	multicolored, A4-ish size, marker on one or both sides	target
construction paper sheets — ambiguous scribbles	multicolored, partially covered in non-representational marks	target
construction paper sheets — blank	white or lightly colored, no marks	target
KEEP bin	clear plastic, blue label sticker, contains crayons and glue stick	tool
wastebasket	small, black plastic, partially filled	tool
wooden art table	low, child-height, marker-stained surface	obstacle
chair	small wooden child's chair, partially on top of fallen paper	obstacle
open backpack	blue nylon, unzipped, paper corner visible inside	distractor
broken crayon pieces	small waxy fragments, mixed colors, on table surface	target
plush elephant	gray fabric, medium-sized, slumped on bookshelf	distractor
glue stick	purple cap, inside KEEP bin	distractor

Expected actions

1. approach the art table from the doorway, navigating around the open backpack on the floor 8s
2. pick up the fallen sheet of paper from under the chair leg, pulling the chair slightly aside to free it 14s
3. rotate the retrieved paper to inspect both sides under available light 6s
4. place the paper into the KEEP bin or wastebasket based on assessed content (drawing vs. blank/scribble) 5s
5. pick up the overhanging papers from the table edge one at a time before they fall, starting with the most precarious 18s
6. rotate each paper in turn to inspect front and back for representational content or the child's name 30s
7. place each inspected paper into KEEP bin or wastebasket according to assessed category 20s
8. pick up remaining papers from the table surface in small stacks of similar-looking sheets 12s
9. separate any stacked pages that are stuck together with dried glue or overlapping marker bleed 10s
10. inspect each separated sheet individually and sort into KEEP bin or wastebasket 35s
11. pick up any loose broken crayon pieces visible on the table surface and place them in the wastebasket 8s
12. verify KEEP bin is upright and contents are lying flat, not crumpled or folded over the bin edge 6s
13. retreat from the table and turn toward the doorway, leaving the backpack and bookshelf items undisturbed 8s

Narration script

00:00 I approach the art table, navigating around the open backpack on the floor — that stays where it is, it's a personal item.

00:08 I can see a sheet of paper partially under the chair leg on the floor. I slide the chair slightly to the side to free it without tearing it.

00:16 I pick up the floor paper and rotate it to inspect both sides. One side has what looks like a blue loop and some orange marks — hard to tell if intentional. I'll treat it as a drawing and place it in the KEEP bin.

00:27 There are papers hanging off the table edge — I pick up the most precarious one first to prevent it from falling. I rotate it: this one has a house with a sun. Clearly a drawing. KEEP bin.

00:38 Next overhanging sheet: this one is about half covered in horizontal green lines, rest blank. Could be a child practicing, could be abandoned. I'm placing it in KEEP — it has deliberate marks.

00:50 I move along the table, picking up remaining surface papers. Two sheets seem stuck together — I separate them carefully by pulling at opposite corners to avoid tearing.

01:02 The top sheet from that pair is completely blank. Wastebasket. The bottom sheet has a stick figure family with names written above them — definitely KEEP.

01:14 Picking up a cluster of three more sheets. One is dense with random overlapping scribble covering the whole page — I'm placing it in the wastebasket. The other two have identifiable shapes and go in the KEEP bin.

01:35 I notice two small broken crayon pieces on the table surface — not inside the KEEP bin. I pick them up and drop them into the wastebasket.

01:48 The remaining sheets on the table include one that's especially ambiguous — half scribble, half what might be an attempt at letters. I flip it over: blank on the back. I choose KEEP, erring toward preserving the child's work.

02:05 Continuing to sort the last few papers. One more blank sheet — wastebasket. The dinosaur drawing and the stick figures both go flat into the KEEP bin.

02:28 The table is now clear. I check the KEEP bin — contents are lying flat, nothing is crumpled or spilling over the edge. The bin is stable and upright.

02:38 I verify I have not touched the backpack, the bookshelf items, or the plush elephant. The wastebasket is not overflowing. The task is complete.

02:48 I retreat from the table and turn toward the doorway. The art table is cleared. What counts as kept versus discarded was genuinely uncertain — that uncertainty is part of the task.

Edge cases

Two sheets are lightly stuck together with dried rubber cement applied by the child — pulling them apart too quickly tears one; the robot must separate them with a slow lateral peel from one corner.
One 'finished' drawing has the child's name written on the back but appears nearly identical to a scribble from the front — the operator must flip every sheet to check the reverse side before discarding.
The KEEP bin tips when a heavy stack of papers is placed off-center, causing already-sorted sheets to spill onto the floor and requiring re-sorting from partial memory of prior decisions.
One sheet of paper contains what appears to be homework with printed text and the child's handwritten answers — it was mixed in with art papers but belongs in the backpack or a separate school pile, introducing a third category the original goal framing did not anticipate.
The ceiling fan's airflow causes lightweight sheets on the table edge to drift or flip over mid-sort, changing what was visible during inspection and requiring the operator to re-examine sheets already thought to be categorized.