Whiteboard affinity mapping
Cluster sticky notes into themes by dragging and label the groups.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Miro · FigJam · Excalidraw · tldraw
- Class GUI-need label
G3· GUI-only
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R11.03.01.02-T01 Cluster 20 feedback stickies into four labeled frames
Original instruction
Open boards/feedback.excalidraw in Excalidraw. Drag the 20 feedback sticky notes into 4 frames by theme (you choose the themes) and title each frame with a 1–3 word theme label. Do not edit note text. Save to output/feedback_clustered.excalidraw.
- Platforms
web· Any OS (web)- Applications
- Excalidraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmenttype· Text & forms- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Parse JSON: each note's frameId; partition vs gold clusters ARI ≥0.8 (graded continuously); 4 frames with names; note text unchanged. Agent has no terminal or text editor and browser DevTools are disabled; element version counters and updated timestamps must show interactive edits, and a single scripted rewrite is detected and scores 0.
Proposed partial credit
15% four frames; 60% clustering ARI; 15% labels present; 10% text untouched
R11.03.01.02-T02 Move all red stickies into the Risks frame
Original instruction
In the Miro board 'Sprint 42' (test team), move every red sticky note into the frame named 'Risks' without changing their text or color. Leave other notes where they are.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropnav· Pan/zoom/scroll- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 1 / 5
Proposed verifier
Verifier REST (agent has no token): all 9 red stickies have parent = Risks frame; non-red stickies' positions unchanged (±1 px); text/color unchanged. Board audit log shows edits from the browser session.
Proposed partial credit
70% red notes in frame; 30% others untouched
R11.03.01.02-T03 Affinity-map 60 interview notes into themes with headers
Original instruction
Miro board 'Onboarding research' has 60 sticky notes from 6 user interviews. Group them into 5–8 themes by dragging notes into clusters (frames or tight groups), add a header sticky above each cluster naming the theme, and leave no note unclustered.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP10· Annotations- Conditions
long· Long-horizon- Proposed difficulty
- 4 / 5
Proposed verifier
Verifier REST: clusters via frame parent or DBSCAN on positions; ARI vs researcher gold ≥0.6 for full credit; every note assigned; header count equals cluster count; header-theme match scored by LLM rubric (flagged non-deterministic). Compared with control pair T17. No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
55% clustering agreement; 15% all notes clustered; 15% headers per cluster; 15% header quality (rubric)
R11.03.01.02-T04 Place stickies on an impact-effort matrix by their tags
Original instruction
In FigJam file 'Backlog triage', a 2x2 matrix (Impact up, Effort right) is drawn. Each of 28 stickies has tags like 'impact:high effort:low' in its text. Drag every sticky into the correct quadrant without overlapping the axis lines.
- Platforms
web· Any OS (web)mac· macOSwin· Windows- Applications
- FigJam
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropgeom· Precise geometry- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Figma REST (verifier token): each sticky bbox fully inside the gold quadrant region (axis margin 8 px); 28/28 scored; text unchanged. FigJam plugins and widgets are disabled for the test org and the agent gets no Figma token (REST cannot write file content), so edits must be made in the UI.
Proposed partial credit
85% quadrant accuracy (about 3 points each); 15% no axis overlap
R11.03.01.02-T05 Stack duplicate notes and cluster the rest
Original instruction
In tldraw (local), open boards/brainstorm.tldr with 45 notes, several of which say the same thing in different words. Stack near-duplicates on top of each other (offset ≤20 px), then cluster stacks and singles into themes with a text label per cluster. Save to output/brainstorm_sorted.tldr.
- Platforms
web· Any OS (web)- Applications
- tldraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmenttype· Text & forms- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Parse .tldr: duplicate groups = notes whose centres within 20 px; precision/recall vs gold duplicate pairs; cluster ARI vs gold ≥0.6; one text label near each cluster. Agent has no terminal or text editor and browser DevTools are disabled, so the .tldr can only be written by the tldraw page.
Proposed partial credit
40% duplicate stacking F1; 40% clustering; 20% labels
R11.03.01.02-T06 Tally dot votes and reorder clusters by score
Original instruction
Miro board 'Retro Oct' has 7 labeled clusters with dot-vote stickers (circles) on notes. Count dots per cluster, write the count into each cluster header as '(n)', and rearrange clusters left to right in descending vote order, keeping each cluster's notes together.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmentdrag· Drag & droptype· Text & forms- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Verifier REST: header texts end with correct '(n)' counts; cluster centroid x order equals descending counts; every note still within its cluster frame/bbox. No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
40% counts correct; 40% order; 20% clusters intact
R11.03.01.02-T07 Color stickies by participant and cluster by theme
Original instruction
In FigJam 'Workshop dump', 36 stickies list their author initials at the end. Recolor each sticky by author (palette in legend frame), then cluster all stickies into themes inside sections named for each theme. Do not change note text.
- Platforms
web· Any OS (web)mac· macOSwin· Windows- Applications
- FigJam
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & dropvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Figma REST: sticky fill colors match author→color legend (36/36); sections contain stickies; ARI vs gold themes ≥0.6; text unchanged. FigJam plugins and widgets are disabled for the test org and the agent gets no Figma token (REST cannot write file content), so edits must be made in the UI.
Proposed partial credit
35% colors; 50% clustering; 15% sections named
R11.03.01.02-T08 Sort research notes into an empathy map
Original instruction
In Mural (test workspace) the template 'Empathy map' has quadrants Says, Thinks, Does, Feels. Drag the 32 interview notes in the parking area into the correct quadrant (quotes go to Says, inferred beliefs to Thinks, observed actions to Does, emotions to Feels).
- Platforms
web· Any OS (web)- Applications
- Mural
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Mural API (verifier): each note's position inside gold quadrant; accuracy over 32 (gold labelled by 2 researchers, ambiguous notes accept either); parking area empty. No Mural API token or OAuth app is issued to the agent and DevTools are disabled, so notes must be dragged in the Mural board UI.
Proposed partial credit
85% quadrant accuracy; 15% parking area cleared
R11.03.01.02-T09 Build a two-level affinity hierarchy with nested frames
Original instruction
Miro board 'Clinic study' has 90 observation notes. Build a two-level hierarchy: 3–4 top-level themes as large frames, each containing 2–4 sub-theme frames with notes inside; label every frame. Every note must sit in exactly one sub-theme.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentnav· Pan/zoom/scroll- Work product
WP10· Annotations- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Verifier REST: frame nesting depth 2; all notes in leaf frames; sub-theme ARI vs gold ≥0.5 and top-level ARI ≥0.6 for full credit; label count = frame count. No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
15% structure valid; 35% sub-theme agreement; 30% top-level agreement; 20% all notes placed and labeled
R11.03.01.02-T10 Draw cause-effect arrows between theme clusters
Original instruction
In output/feedback_clustered.excalidraw (5 labeled clusters), add arrows bound to cluster frames showing causal links stated in causal_notes.md (e.g. 'Slow onboarding → churn'), each arrow labeled with the evidence note id. Do not move notes.
- Platforms
web· Any OS (web)- Applications
- Excalidraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & droptype· Text & formspath· Continuous path- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Parse JSON: arrows with startBinding/endBinding to the correct frames for each of 6 gold links; labels contain note ids; no extra links; note positions unchanged. Agent has no terminal or text editor and browser DevTools are disabled, so the .excalidraw JSON can only be written by the Excalidraw page.
Proposed partial credit
60% correct bound links; 25% labels; 15% notes unmoved
R11.03.01.02-T11 Merge a second team's clusters into the main board
Original instruction
Two teams clustered the same research separately. Copy every note from FigJam file 'Team B board' into 'Team A board', placing each into Team A's matching theme section; create new sections only for themes Team A lacks. Merge exact-duplicate notes.
- Platforms
web· Any OS (web)mac· macOSwin· Windows- Applications
- FigJam
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropmulti· Cross-appvis· Visual judgment- Work product
WP10· Annotations- Conditions
multiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Figma REST: Team A contains union of notes (exact duplicates once); each imported note in gold-mapped section; new sections only for 2 missing themes; Team B unchanged. FigJam plugins and widgets are disabled for the test org and the agent gets no Figma token (REST cannot write file content), so edits must be made in the UI.
Proposed partial credit
30% notes copied; 40% placement; 15% duplicates merged; 15% correct new sections
R11.03.01.02-T12 Circle the legacy clusters with freehand strokes
Original instruction
In tldraw open boards/arch_notes.tldr. Draw a freehand loop around each of the 3 clusters whose header contains 'Legacy' (each loop must enclose all notes of that cluster and no other notes). Save to output/arch_marked.tldr.
- Platforms
web· Any OS (web)- Applications
- tldraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
path· Continuous pathvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Parse .tldr: 3 draw shapes whose closed polygons contain all notes of each Legacy cluster (point-in-polygon on note centres) and none from other clusters. Agent has no terminal or text editor and browser DevTools are disabled, so the .tldr can only be written by the tldraw page.
Proposed partial credit
30% three loops drawn; 50% containment correct; 20% no extra notes enclosed
R11.03.01.02-T13 Use auto-clustering by tag, then fix and rename clusters
Original instruction
In Miro board 'Feature requests', select the 50 stickies and use 'Cluster by keywords/tags' to group them, then fix 6 mis-clustered notes the tags got wrong (per their text) and rename each cluster header to a theme name.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & dropvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Verifier REST: cluster partition ARI vs gold ≥0.85 (auto-cluster alone scores ~0.7); the 6 mis-tagged notes in gold clusters; headers renamed (not default tag names). No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
30% auto-cluster used; 45% corrections; 25% headers renamed
R11.03.01.02-T14 Group stickies into labeled clusters in Freeform
Original instruction
In Apple Freeform board 'Trip ideas' (24 sticky notes), group the notes into 4 clusters (Food, Outdoors, Culture, Logistics) placed in the four corners of the board, each with a text label above, and save.
- Platforms
mac· macOSios· iOS/iPadOS- Applications
- Freeform
- Interfaces
desk· Desktop GUImob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & droptouch· Touch gesturesvis· Visual judgment- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Board exported as PDF and via the Freeform SQLite store: note positions per quadrant equal gold cluster; 4 labels positioned above clusters; text unchanged. The agent has no terminal and Shortcuts/AppleScript automation is denied by TCC; writes to Freeform's store by any other process are caught by an audit log.
Proposed partial credit
75% notes in correct corner cluster; 25% labels
R11.03.01.02-T15 Run a retro board: sort, merge, vote and create actions
Original instruction
Miro retro board 'Sprint 43': move 40 loose notes into Start/Stop/Continue columns, stack duplicates, apply the existing votes (stickers) to find the top 3 items, and create a Miro card for each in frame 'Actions' with an owner from owners.txt and due date next Friday.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropclick· Point & clicktype· Text & forms- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Verifier REST: column placement accuracy vs gold; duplicate stacks; Actions frame holds 3 cards whose titles match top-3 voted notes, assignee and due date correct. No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
40% column placement; 15% duplicates; 30% top-3 actions; 15% owners/dates
R11.03.01.02-T16 Lay out a journey map across stages and persona lanes
Original instruction
FigJam file 'Journey' has 6 stage columns (Discover → Advocate) and 2 persona swimlanes (Clinician, Patient). Place 40 loose stickies into the right stage-lane cell, ordered left-to-right within each cell by the step number in the note, without overlaps.
- Platforms
web· Any OS (web)mac· macOSwin· Windows- Applications
- FigJam
- Interfaces
brow· Browserdesk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropgeom· Precise geometryvis· Visual judgment- Work product
WP10· Annotations- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Figma REST: each sticky bbox inside gold cell; within-cell x order matches step numbers (Kendall tau = 1 per cell); no pairwise overlaps. FigJam plugins and widgets are disabled for the test org and the agent gets no Figma token (REST cannot write file content), so edits must be made in the UI.
Proposed partial credit
50% cell accuracy; 30% ordering; 20% no overlaps
R11.03.01.02-T17 Cluster the interview notes via the Miro REST API (control pair of T03)
Original instruction
Control pair of T03. With a Miro REST test token, write cluster.py that reads the 60 notes, assigns themes (you may embed and cluster their text), creates frames with header stickies, and moves notes into them.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Miro REST API
- Interfaces
api· API/SDK- GUI need · task setting
AP· API-sufficient- GUI need · ordinary access
AP· API-sufficient- Interaction primitives
code· Code & commands- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Same scoring as T03 on final board state; audit shows edits from the token's app.
Proposed partial credit
55% clustering; 15% all notes; 15% headers; 15% header quality
R11.03.01.02-T18 Label pre-built clusters and add a summary note to each
Original instruction
In Excalidraw boards/usability.excalidraw, notes are already grouped into 6 unlabeled clusters. Add a 2–4 word title above each cluster and a one-sentence summary sticky (≤20 words) beside it that names the main problem and how many notes support it.
- Platforms
web· Any OS (web)- Applications
- Excalidraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsgeom· Precise geometry- Work product
WP10· Annotations- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Parse JSON: 6 titles positioned above clusters, 6 summaries adjacent; supporting-note counts in summaries exact; title/summary relevance scored by LLM rubric against gold themes (flagged non-deterministic). Agent has no terminal or text editor and browser DevTools are disabled, so the .excalidraw JSON can only be written by the Excalidraw page.
Proposed partial credit
20% placement; 30% counts correct; 50% title/summary quality (rubric)
R11.03.01.02-T19 Cluster 150 scattered notes on a large board
Original instruction
Miro board 'Citizen feedback' has 150 notes scattered across a very large canvas (some far off-screen) from 3 workshops. Gather all of them into 8 themed clusters inside a new frame 'Synthesis' with headers, keeping a copy count per cluster in the header '(n)'.
- Platforms
web· Any OS (web)- Applications
- Miro
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropnav· Pan/zoom/scrollvis· Visual judgment- Work product
WP10· Annotations- Conditions
long· Long-horizonhidden· Hidden/stale state- Proposed difficulty
- 5 / 5
Proposed verifier
Verifier REST: all 150 notes inside Synthesis (no stragglers); ARI vs gold ≥0.5 for full credit; header count 8 with correct '(n)' counts matching actual cluster sizes. No Miro API token or app install is given to the agent (the verifier token is grader-only) and DevTools are disabled, so notes must be moved in the board UI.
Proposed partial credit
25% all notes gathered; 45% clustering; 15% headers; 15% counts
R11.03.01.02-T20 Keep clustering while a teammate adds notes live
Original instruction
In a shared tldraw room, cluster the 30 existing notes into themes with labels. A scripted collaborator keeps adding a new note every 30–60 s for 8 minutes; place each new note into the right cluster (or a new one) within 2 minutes of its arrival.
- Platforms
web· Any OS (web)- Applications
- tldraw
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentrt· Real-time control- Work product
WP10· Annotations- Conditions
interrupt· Interruptions- Proposed difficulty
- 4 / 5
Proposed verifier
Room event log: initial clustering ARI vs gold ≥0.6; for each of ~10 new notes, time from creation to placement in gold cluster ≤120 s; final state all notes clustered. Browser DevTools are disabled and the agent has no terminal or sync-server access; the room event log attributes each move to the agent's browser session.
Proposed partial credit
40% initial clustering; 40% new notes placed in time; 20% final completeness
Mapped research references
No benchmark reference is assigned to this class in the current mapping.